EAGLES Synopsis and Comparison of Morphosyntactic

Phenomena Encoded in Lexicons and Corpora. A Common Proposal and Applications to European

Languages

EAGLES Document EAG-CLWG-MORPHSYN/R

Version of May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R Contents 1 Authors 17 2 Other Contributors 17 3 Commentators 17 4 Introduction 17

4.1 Standards and Shareable Resources: the Area of Morphosyntax . . . . . . . . . . 17

5 Methodolgogy Adopted 18

5.1 Interdependence between the Lexicon and the Corpus View . . . . . . . . . . . . 18 5.2 Bottom-up approach . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19 5.3 The Survey and Comparison Phase . . . . . . . . . . . . . . . . . . . . . . . . . . 19

5.3.1 Description of the first zone of the tables . . . . . . . . . . . . . . . . . . 20 5.4 The Proposal Phase . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21

5.4.1 Description of the second zone of the tables . . . . . . . . . . . . . . . . . 21 5.5 The Application Phase . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22

5.5.1 Description of the Tables used in the Applications . . . . . . . . . . . . . 22 5.6 Interaction with some European Projects . . . . . . . . . . . . . . . . . . . . . . 23

5.6.1 The LRE-MULTEXT Experience . . . . . . . . . . . . . . . . . . . . . . . 23 5.6.2 The MLAP-PAROLE Experience . . . . . . . . . . . . . . . . . . . . . . . 25 5.7 The Guidelines of the Validation Phase . . . . . . . . . . . . . . . . . . . . . . . 25

6 Some relevant aspects for lexicon specifications 26

6.1 Lexical specifications vs. corpus tagsets . . . . . . . . . . . . . . . . . . . . . . . 26 6.2 The concept of "common" category . . . . . . . . . . . . . . . . . . . . . . . . . . 28 6.3 Objects described and level of description . . . . . . . . . . . . . . . . . . . . . . 28 6.4 Descriptive approach over different languages wrt monotonicity . . . . . . . . . . 29

7 Language-specific Applications 30

7.1 Application to Italian . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30 7.2 Application to German . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30 7.3 Application to English . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31 7.4 Application to Dutch . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31 7.5 Application to Spanish . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32 7.6 Application to French . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32 7.7 Application to Portuguese . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33 7.8 Application to Danish . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33 7.9 Application to Greek . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33 7.10 Application to Swedish . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35 7.11 Application to Catalan . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35 7.12 Application to Irish . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35

8 Present Status of the Morphosyntactic Proposal 35

Recommendations 1 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R 9 Noun 37

9.1 Comments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38

9.1.1 Type . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38 9.1.2 Number and Gender: The "common" and "invariant" values . . . . . . . 38 9.1.3 Case . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38 9.1.4 Countability . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39 9.1.5 Values on Level 2b - language specific . . . . . . . . . . . . . . . . . . . . 39 9.2 Application to Italian (Dictionary and Corpus) . . . . . . . . . . . . . . . . . . . 40

9.2.1 Type . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40 9.2.2 Gender . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40 9.2.3 Number . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40 9.2.4 Countability . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41 9.2.5 Features not pertinent to Italian . . . . . . . . . . . . . . . . . . . . . . . 41 9.3 Application to German . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42

9.3.1 Type . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42 9.3.2 Gender . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42 9.3.3 Number . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42 9.3.4 Countability . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42 9.3.5 Case . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42 9.3.6 language-specific features . . . . . . . . . . . . . . . . . . . . . . . . . . . 42 9.4 Application to English . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42

9.4.1 Type . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43 9.4.2 Number . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43 9.5 Application to Dutch . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44

9.5.1 Type . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44 9.5.2 Gender . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44 9.5.3 Number . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45 9.5.4 Countability . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45 9.5.5 Case . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46 9.5.6 Language-specific features . . . . . . . . . . . . . . . . . . . . . . . . . . . 46 9.6 Application to Spanish . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47

9.6.1 Type . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47 9.6.2 Gender . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47 9.6.3 Number . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48 9.6.4 Case . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48 9.6.5 Countability . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48 9.7 Application to French (Corpus) . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49

9.7.1 Type . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49 9.7.2 Gender . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49 9.7.3 Number . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49 9.7.4 EAGLES features not applicable . . . . . . . . . . . . . . . . . . . . . . . 49 9.7.5 IBMF Tagset features not applicable in EAGLES . . . . . . . . . . . . . . 49 9.8 Application to French (Lexicon) . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50

9.8.1 Type . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50 9.8.2 Gender . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50 9.8.3 Number . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50 9.8.4 Case . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50

Recommendations 2 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R

9.8.5 Combinations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50 9.9 Application to Portuguese . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51

9.9.1 Type . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51 9.9.2 Gender . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51 9.9.3 Number . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51 9.9.4 Countability . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52 9.9.5 Features not pertinent to Portuguese . . . . . . . . . . . . . . . . . . . . . 52 9.10 Application to Danish . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53

9.10.1 Type . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53 9.10.2 Case . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53 9.10.3 Countability . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53 9.10.4 Gender . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53 9.10.5 Number . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54 9.10.6 Language-specific features: Definiteness . . . . . . . . . . . . . . . . . . . 54 9.11 Application to Greek . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55

9.11.1 Type . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55 9.11.2 Gender . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55 9.11.3 Number . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55 9.11.4 Case . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56 9.11.5 Countability . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57 9.11.6 Features not pertinent to Greek . . . . . . . . . . . . . . . . . . . . . . . . 57 9.12 Application to Swedish . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58

9.12.1 Type . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58 9.12.2 Gender . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58 9.12.3 Number . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58 9.12.4 Definiteness . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59 9.12.5 Case . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59 9.13 Application to Catalan . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60

9.13.1 Type . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60 9.13.2 Gender . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60 9.13.3 Number . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60 9.13.4 Countability . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61 9.13.5 Features not pertinent to Catalan . . . . . . . . . . . . . . . . . . . . . . 61 9.14 Application to Irish . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62

9.14.1 Type . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62 9.14.2 Gender . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63 9.14.3 Number . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63 9.14.4 Case . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63 9.14.5 Features not pertinent to Irish . . . . . . . . . . . . . . . . . . . . . . . . 63 9.14.6 Language specific features: Contrast . . . . . . . . . . . . . . . . . . . . . 63

10 Verb 65

10.1 Comments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66

10.1.1 Type . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66 10.1.2 Finiteness, Verb-Form/Mood . . . . . . . . . . . . . . . . . . . . . . . . . 66 10.1.3 Tense . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67 10.1.4 Person, Gender and Number . . . . . . . . . . . . . . . . . . . . . . . . . 67

Recommendations 3 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R

10.1.5 Aspect . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68 10.1.6 Voice . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68 10.1.7 Reflexivity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68 10.1.8 Main-Verb Function and Auxiliary Function . . . . . . . . . . . . . . . . . 68 10.1.9 Level-2b specifications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68 10.2 Application to Italian . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70

10.2.1 Type . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70 10.2.2 Finiteness and Verb-Form/Mood . . . . . . . . . . . . . . . . . . . . . . . 70 10.2.3 Tense . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70 10.2.4 Person . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71 10.2.5 Gender . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71 10.2.6 Number . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72 10.2.7 Combination of features . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72 10.2.8 Enclitic Phenomenon . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73 10.3 Application to German . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 74

10.3.1 Type . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 74 10.3.2 Finiteness, Verb-Form/Mood . . . . . . . . . . . . . . . . . . . . . . . . . 74 10.3.3 Tense . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 74 10.3.4 Person . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75 10.3.5 Number . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75 10.3.6 Gender . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75 10.4 Application to English . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 76

10.4.1 Type . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 76 10.4.2 Finiteness . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 76 10.4.3 Verb-form . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77 10.4.4 Mood . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77 10.4.5 Tense . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77 10.4.6 Person . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77 10.4.7 Number . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77 10.4.8 Auxiliary Type . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78 10.5 Application to Dutch . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79

10.5.1 Verb-Type . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79 10.5.2 Verb-Form . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79 10.5.3 Mood . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80 10.5.4 Tense . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81 10.5.5 Person . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81 10.5.6 Number . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81 10.5.7 Gender . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82 10.5.8 Main-Verb Function . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82 10.5.9 Auxiliary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82 10.5.10 Other Verb-Form features . . . . . . . . . . . . . . . . . . . . . . . . . . . 82 10.5.11 Separability . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82 10.5.12 Inversion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 83 10.5.13 Word order separable verbs . . . . . . . . . . . . . . . . . . . . . . . . . . 83 10.5.14 Politeness . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 84 10.6 Application to Spanish . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85

10.6.1 Verb-Form and Mood . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85

Recommendations 4 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R

10.6.2 Tense . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 86 10.6.3 Person . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 86 10.6.4 Number . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 86 10.6.5 Gender . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 86 10.7 Application to French (Corpus) . . . . . . . . . . . . . . . . . . . . . . . . . . . . 87

10.7.1 Type . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 87 10.7.2 Finiteness and Verb-Form/Mood . . . . . . . . . . . . . . . . . . . . . . . 87 10.7.3 Tense . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 87 10.7.4 Person . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 87 10.7.5 Gender . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 88 10.7.6 Number . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 88 10.7.7 EAGLES features not applicable . . . . . . . . . . . . . . . . . . . . . . . 88 10.7.8 IBMF Tagset features not applicable in EAGLES . . . . . . . . . . . . . . 88 10.8 Application to French (Lexicon) . . . . . . . . . . . . . . . . . . . . . . . . . . . . 89

10.8.1 Type . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 89 10.8.2 Finiteness . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 89 10.8.3 Verb Form/Mood . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 89 10.8.4 Tense . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 89 10.8.5 Person . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 90 10.8.6 Number . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 90 10.8.7 Gender . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 90 10.8.8 Combinations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 90 10.9 Application to Portuguese . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 94

10.9.1 Type . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 94 10.9.2 Finiteness . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 94 10.9.3 Mood . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 94 10.9.4 Tense . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 95 10.9.5 Person . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 95 10.9.6 Gender . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 95 10.9.7 Number . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 96 10.9.8 Combination of features . . . . . . . . . . . . . . . . . . . . . . . . . . . . 97 10.9.9 Clitics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 99 10.10Application to Danish . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 100

10.10.1 Type . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 100 10.10.2 Finiteness . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 100 10.10.3 Verb-form . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 100 10.10.4 Mood . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 101 10.10.5 Tense . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 101 10.10.6 Number . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 101 10.10.7 Person . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 101 10.10.8 Gender . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 101 10.10.9 language-specific feature: Voice . . . . . . . . . . . . . . . . . . . . . . . . 101 10.11Application to Greek . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 103

10.11.1 Type . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 103 10.11.2 Finiteness and Verb-Form/Mood . . . . . . . . . . . . . . . . . . . . . . . 104 10.11.3 Tense . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 105 10.11.4 Person . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 105

Recommendations 5 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R

10.11.5 Number . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 106 10.11.6 Gender and Case . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 106 10.11.7 Aspect . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 108 10.11.8 Voice . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 109 10.12Application to Swedish . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 110

10.12.1 Type . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 110 10.12.2 Form/Mood . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 110 10.12.3 Tense . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 111 10.12.4 Voice . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 111 10.12.5 Gender/Number . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 111 10.12.6 Person . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 111 10.13Application to Catalan . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 112

10.13.1 Type . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 112 10.13.2 Finiteness and Verb-Form/Mood . . . . . . . . . . . . . . . . . . . . . . . 112 10.13.3 Tense . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 113 10.13.4 Person . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 113 10.13.5 Gender . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 115 10.14Application to Irish . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 116

10.14.1 Type . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 116 10.14.2 Finiteness . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 116 10.14.3 Mood . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 116 10.14.4 Tense . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 117 10.14.5 Person . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 117 10.14.6 Gender, Aspect, Voice, etc. . . . . . . . . . . . . . . . . . . . . . . . . . . 117 10.14.7 Language-specific features: The Dependent Form . . . . . . . . . . . . . . 117

11 Adjective 119

11.1 Comments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 120

11.1.1 Type . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 120 11.1.2 Degree . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 120 11.1.3 Gender, Number and Case . . . . . . . . . . . . . . . . . . . . . . . . . . . 121 11.1.4 Use . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 121 11.1.5 Modification Type . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 121 11.1.6 Inflection Type . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 121 11.1.7 Definiteness . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 121 11.2 Application to Italian . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 122

11.2.1 Type . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 122 11.2.2 Degree . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 122 11.2.3 Gender and Number . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 122 11.2.4 Use . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 123 11.2.5 Modification Type . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 123 11.2.6 Features not applicable in Italian . . . . . . . . . . . . . . . . . . . . . . . 124 11.3 Application to German . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 125

11.3.1 Degree of comparison . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 125 11.3.2 Gender . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 125 11.3.3 Number . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 125 11.3.4 Case . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 125

Recommendations 6 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R

11.3.5 Use . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 125 11.3.6 Inflection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 126 11.4 Application to English . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 127

11.4.1 Degree . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 127 11.5 Application to Dutch . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 128

11.5.1 Degree of comparison . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 128 11.5.2 Gender . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 128 11.5.3 Number . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 128 11.5.4 Case . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 128 11.5.5 Use . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 128 11.5.6 Inflection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 129 11.5.7 Language-specific feature: non-verbal part of separable verb . . . . . . . . 130 11.6 Application to Spanish . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 131

11.6.1 Type . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 131 11.6.2 Degree . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 131 11.6.3 Gender . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 131 11.6.4 Number . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 131 11.6.5 Modification Type . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 131 11.7 Application to French (Corpus) . . . . . . . . . . . . . . . . . . . . . . . . . . . . 133

11.7.1 Type . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 133 11.7.2 Gender . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 133 11.7.3 Number . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 133 11.7.4 EAGLES features not applicable . . . . . . . . . . . . . . . . . . . . . . . 133 11.7.5 IBMF Tagset features not applicable in EAGLES . . . . . . . . . . . . . . 133 11.8 Application to French (Lexicon) . . . . . . . . . . . . . . . . . . . . . . . . . . . . 133

11.8.1 Type . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 133 11.8.2 Degree . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 134 11.8.3 Gender . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 135 11.8.4 Number . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 135 11.8.5 Case . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 135 11.8.6 Combinations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 135 11.9 Application to Portuguese . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 137

11.9.1 Type . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 137 11.9.2 Degree . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 137 11.9.3 Gender . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 137 11.9.4 Number . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 137 11.9.5 Case . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 138 11.9.6 Other features required to encode Possessive Adjectives . . . . . . . . . . 138 11.9.7 Possessor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 138 11.9.8 Person . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 138 11.10Application to Danish . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 139

11.10.1 Type . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 139 11.10.2 Degree . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 139 11.10.3 Gender, Number and Definiteness . . . . . . . . . . . . . . . . . . . . . . 139 11.10.4 Use . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 140 11.10.5 Synoptic table of agreement relations . . . . . . . . . . . . . . . . . . . . . 140 11.11Application to Greek . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 142

Recommendations 7 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R

11.11.1 Degree . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 142 11.11.2 Gender, Number and Case . . . . . . . . . . . . . . . . . . . . . . . . . . . 143 11.11.3 Use . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 144 11.11.4 Comments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 144 11.11.5 Features not applicable in Greek . . . . . . . . . . . . . . . . . . . . . . . 144 11.12Application to Swedish . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 145

11.12.1 Type . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 145 11.12.2 Use . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 145 11.12.3 Degree . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 145 11.12.4 Definiteness . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 146 11.12.5 Number . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 146 11.12.6 Gender . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 146 11.12.7 Case . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 146 11.13Application to Catalan . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 148

11.13.1 Type . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 148 11.13.2 Degree . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 148 11.13.3 Gender . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 148 11.13.4 Number . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 149 11.14Application to Irish . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 150

11.14.1 Degree . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 150 11.14.2 Gender . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 150 11.14.3 Number . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 150 11.14.4 Number . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 150 11.14.5 Other attributes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 151

12 Pronoun 152

12.1 Comments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 153

12.1.1 Type, Wh-Type . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 153 12.1.2 Person, Case and Possessor . . . . . . . . . . . . . . . . . . . . . . . . . . 153 12.1.3 Politeness . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 154 12.1.4 Function . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 154 12.1.5 Other values on Level 2b . . . . . . . . . . . . . . . . . . . . . . . . . . . 154 12.2 Application to Italian . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 155

12.2.1 Type . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 155 12.2.2 Personal Pronoun . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 155 12.2.3 Reflexive (Pronoun) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 157 12.2.4 Possessive . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 157 12.2.5 Possessor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 158 12.2.6 Demonstrative . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 158 12.2.7 Indefinite . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 158 12.2.8 Interrogative . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 159 12.2.9 Exclamatory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 159 12.2.10 Relatives . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 159 12.3 Application to German . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 160

12.3.1 Type . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 160 12.3.2 Gender . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 160 12.3.3 Number . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 160

Recommendations 8 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R

12.3.4 Case . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 160 12.3.5 Possessor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 160 12.4 Application to English . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 161

12.4.1 Pronoun-type . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 161 12.4.2 Wh-subtype . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 161 12.4.3 Person . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 161 12.4.4 Number . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 161 12.4.5 Gender . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 161 12.4.6 Case . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 162 12.5 Application to Dutch . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 163

12.5.1 Type . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 163 12.5.2 Gender . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 163 12.5.3 Number . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 163 12.5.4 Case . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 164 12.5.5 Function . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 164 12.6 Application to Spanish . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 165

12.6.1 Type . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 165 12.6.2 WH-Type . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 166 12.7 Application to French (Corpus) . . . . . . . . . . . . . . . . . . . . . . . . . . . . 168

12.7.1 Type . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 168 12.7.2 wh-Subtype . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 168 12.7.3 Person . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 168 12.7.4 Gender . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 168 12.7.5 Number . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 168 12.7.6 Case . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 169 12.7.7 EAGLES features not applicable . . . . . . . . . . . . . . . . . . . . . . . 170 12.8 Application to French (Lexicon) . . . . . . . . . . . . . . . . . . . . . . . . . . . . 171

12.8.1 Type . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 171 12.8.2 Person . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 171 12.8.3 Gender . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 171 12.8.4 Number . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 171 12.8.5 Case . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 172 12.8.6 Possessor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 172 12.8.7 Combinations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 172 12.9 Application to Portuguese . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 175

12.9.1 Type . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 175 12.9.2 Person . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 175 12.9.3 Gender . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 175 12.9.4 Number . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 175 12.9.5 Case . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 176 12.10Application to Danish . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 177

12.10.1 Type . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 177 12.10.2 Wh-subtype . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 177 12.10.3 Gender . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 177 12.10.4 Number . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 178 12.10.5 Case . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 178 12.10.6 Politeness . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 178

Recommendations 9 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R

12.10.7 Function . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 178 12.11Application to Greek . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 179

12.11.1 Pronoun Types . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 179 12.11.2 Personal Pronouns . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 180 12.11.3 Politeness form . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 180 12.11.4 Reflexive Pronoun . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 181 12.11.5 Possessive pronouns . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 181 12.11.6 Demonstrative pronouns . . . . . . . . . . . . . . . . . . . . . . . . . . . . 182 12.11.7 Indefinite pronouns . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 183 12.11.8 Interrogative Pronouns . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 183 12.11.9 Relative pronouns . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 184 12.11.10Relative-Indefinite Pronouns . . . . . . . . . . . . . . . . . . . . . . . . . 184 12.11.11Definite Pronouns . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 184 12.11.12Clitic Pronouns . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 185 12.12Application to Swedish . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 187

12.12.1 Type . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 187 12.12.2 Number . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 187 12.12.3 Person . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 187 12.12.4 Gender . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 188 12.12.5 Case . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 188 12.13Application to Catalan . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 189 12.14Application to Irish . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 190

12.14.1 Type and Wh-type . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 190 12.14.2 Person . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 190 12.14.3 Gender . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 190 12.14.4 Number . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 190 12.14.5 EAGLES features that do not apply . . . . . . . . . . . . . . . . . . . . . 191

13 Determiner 192

13.1 Comments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 193 13.2 Application to Italian . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 194

13.2.1 Type and wh-Type . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 194 13.2.2 Possessive . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 194 13.2.3 Demonstrative . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 195 13.2.4 Indefinite . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 195 13.2.5 Interrogative . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 195 13.2.6 Exclamatory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 196 13.2.7 Relatives . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 196 13.3 Application to German . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 197

13.3.1 Type . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 197 13.3.2 Gender . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 197 13.3.3 Number . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 197 13.3.4 Case . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 197 13.3.5 Possessor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 197 13.4 Application to English . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 198

13.4.1 Type . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 198 13.4.2 Person . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 198

Recommendations 10 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R

13.4.3 Gender . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 198 13.4.4 Number . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 198 13.5 Application to Spanish . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 199 13.6 Application to French (Corpus) . . . . . . . . . . . . . . . . . . . . . . . . . . . . 200

13.6.1 Determiner type . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 200 13.7 Application to French (Lexicon) . . . . . . . . . . . . . . . . . . . . . . . . . . . . 201

13.7.1 Det.-Type . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 201 13.7.2 Person . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 201 13.7.3 Gender . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 201 13.7.4 Number . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 201 13.7.5 Possessor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 201 13.7.6 Combinations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 202 13.8 Application to Portuguese . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 204

13.8.1 Type . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 204 13.8.2 Wh-Type . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 204 13.8.3 Person . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 204 13.8.4 Gender . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 204 13.8.5 Number . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 204 13.8.6 Case . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 204 13.8.7 Possessor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 205 13.9 Application to Danish . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 206

13.9.1 Determiner type . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 206 13.10Application to Greek . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 207 13.11Application to Swedish . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 208 13.12Application to Catalan . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 209 13.13Application to Irish . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 210

13.13.1 Type . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 210 13.13.2 Person, Gender, Number, Case and Possessor . . . . . . . . . . . . . . . . 210

14 Article 211

14.1 Comments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 212 14.2 Application to Italian . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 213

14.2.1 Article . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 213 14.3 Application to German . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 214

14.3.1 Types . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 214 14.3.2 Gender . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 214 14.3.3 Number . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 214 14.3.4 Case . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 214 14.4 Application to English . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 215

14.4.1 Type . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 215 14.4.2 Number . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 215 14.5 Application to Spanish . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 216

14.5.1 Article Type . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 216 14.5.2 Gender . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 216 14.5.3 Number . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 216 14.5.4 Case . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 216 14.6 Application to French (Corpus) . . . . . . . . . . . . . . . . . . . . . . . . . . . . 217

Recommendations 11 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R

14.6.1 Definiteness . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 217 14.6.2 Gender . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 217 14.6.3 Number . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 217 14.6.4 EAGLES features not applicable . . . . . . . . . . . . . . . . . . . . . . . 217 14.7 Application to French (Lexicon) . . . . . . . . . . . . . . . . . . . . . . . . . . . . 217

14.7.1 Type . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 217 14.7.2 Gender . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 217 14.7.3 Number . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 217 14.7.4 Case . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 218 14.7.5 Combinations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 218 14.8 Application to Danish . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 219

14.8.1 Type . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 219 14.8.2 Language-specific property . . . . . . . . . . . . . . . . . . . . . . . . . . 219 14.9 Application to Greek . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 220

14.9.1 Gender, Number and Case . . . . . . . . . . . . . . . . . . . . . . . . . . . 220 14.10Application to Swedish . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 222

14.10.1 Type . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 222 14.10.2 Gender . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 222 14.11Application to Catalan . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 223 14.12Application to Irish . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 224

15 Adverb 225

15.1 Comments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 226

15.1.1 Type . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 226 15.1.2 Degree . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 226 15.1.3 Polarity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 226 15.1.4 Other Distinctions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 226 15.2 Application to Italian . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 227

15.2.1 Degree . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 227 15.3 Application to German . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 228

15.3.1 Type . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 228 15.3.2 Degree of comparison . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 228 15.4 Application to English . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 229

15.4.1 Adverb-type . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 229 15.4.2 Degree . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 229 15.4.3 Polarity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 229 15.4.4 Wh-subtype . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 229 15.5 Application to Dutch . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 230

15.5.1 Type: Language-specific features . . . . . . . . . . . . . . . . . . . . . . . 230 15.5.2 Degree of comparison . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 230 15.6 Application to Spanish . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 231

15.6.1 Type (semantic) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 231 15.6.2 Degree . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 231 15.6.3 Polarity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 232 15.6.4 Apocope . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 232 15.7 Application to French (Corpus) . . . . . . . . . . . . . . . . . . . . . . . . . . . . 233

15.7.1 EAGLES features not applicable . . . . . . . . . . . . . . . . . . . . . . . 233

Recommendations 12 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R

15.7.2 IBMF Tagset features not applicable in EAGLES . . . . . . . . . . . . . . 233 15.8 Application to French (Lexicon) . . . . . . . . . . . . . . . . . . . . . . . . . . . . 233

15.8.1 Type . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 233 15.8.2 Degree . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 233 15.8.3 Combinations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 233 15.9 Application to Portuguese . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 235

15.9.1 Type . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 235 15.9.2 Degree . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 235 15.10Application to Danish . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 236

15.10.1 Degree . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 236 15.11Application to Greek . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 237

15.11.1 Type . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 237 15.11.2 Degree . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 237 15.11.3 Function . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 237 15.11.4 Features not applicable in Greek . . . . . . . . . . . . . . . . . . . . . . . 239 15.12Application to Swedish . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 240

15.12.1 Degree . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 240 15.13Application to Catalan . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 241 15.14Application to Irish . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 242

15.14.1 Type . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 242 15.14.2 Degree . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 242

16 Adposition 243

16.1 Comments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 244

16.1.1 Type . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 244 16.1.2 Formation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 244 16.2 Application to Italian . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 245

16.2.1 Type . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 245 16.2.2 Formation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 245 16.3 Application to German . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 246

16.3.1 Type . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 246 16.3.2 Formation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 246 16.4 Application to English . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 247

16.4.1 Adposition-type . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 247 16.5 Application to Dutch . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 248

16.5.1 Type . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 248 16.6 Application to Spanish . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 249

16.6.1 Prepositions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 249 16.7 Application to French (Corpus) . . . . . . . . . . . . . . . . . . . . . . . . . . . . 250 16.8 Application to French (Lexicon) . . . . . . . . . . . . . . . . . . . . . . . . . . . . 250

16.8.1 Type . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 250 16.8.2 Combinations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 250 16.9 Application to Portuguese . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 251

16.9.1 Type . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 251 16.9.2 Formation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 251 16.10Application to Danish . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 252

16.10.1 Type . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 252

Recommendations 13 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R

16.11Application to Greek . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 253

16.11.1 Type . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 253 16.11.2 Form, Gender, Number and Case . . . . . . . . . . . . . . . . . . . . . . . 253 16.11.3 Function . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 254 16.12Application to Swedish . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 256

16.12.1 Type . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 256 16.13Application to Catalan . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 257 16.14Application to Irish . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 258

16.14.1 Person . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 258 16.14.2 Number . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 258 16.14.3 Gender . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 258

17 Conjunction 259

17.1 Comments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 260

17.1.1 Coordination Type . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 260 17.1.2 Subordination Type . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 260 17.2 Application to Italian . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 261

17.2.1 Type . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 261 17.3 Application to German . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 262

17.3.1 Type . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 262 17.3.2 Subord-Type . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 262 17.4 Application to English . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 263

17.4.1 Conjunction-type . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 263 17.4.2 Coord.-type . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 263 17.4.3 Subord.-type . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 263 17.5 Application to Dutch . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 264 17.6 Application to Spanish . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 265

17.6.1 Type . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 265 17.6.2 Coord-Type . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 265 17.6.3 Subord-Type . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 266 17.7 Application to French (Corpus) . . . . . . . . . . . . . . . . . . . . . . . . . . . . 267 17.8 Application to French (Lexicon) . . . . . . . . . . . . . . . . . . . . . . . . . . . . 267

17.8.1 Type . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 267 17.8.2 Combinations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 267 17.9 Application to Portuguese . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 268

17.9.1 Type . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 268 17.10Application to Danish . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 269

17.10.1 Type . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 269 17.11Application to Greek . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 270

17.11.1 Type . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 270 17.11.2 Coordination Type . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 270 17.12Application to Swedish . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 271

17.12.1 Conj.Type . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 271 17.12.2 Coord.Type . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 271 17.13Application to Catalan . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 272 17.14Application to Irish . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 273

17.14.1 Type . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 273

Recommendations 14 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R

17.14.2 Language specific features . . . . . . . . . . . . . . . . . . . . . . . . . . . 273 18 Numeral 274

18.1 Comments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 275

18.1.1 Type . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 275 18.1.2 Function . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 275 18.2 Application to Italian . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 276

18.2.1 Type . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 276 18.2.2 Gender . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 276 18.2.3 Number . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 276 18.2.4 Function . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 276 18.3 Application to German . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 277

18.3.1 Numerals: Type . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 277 18.4 Application to English . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 278

18.4.1 Numeral-type . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 278 18.4.2 Number . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 278 18.5 Application to Dutch . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 279

18.5.1 Type . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 279 18.6 Application to French (Corpus) . . . . . . . . . . . . . . . . . . . . . . . . . . . . 280 18.7 Application to French (Lexicon) . . . . . . . . . . . . . . . . . . . . . . . . . . . . 280

18.7.1 Type . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 280 18.7.2 Gender . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 280 18.7.3 Number . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 280 18.7.4 Case . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 280 18.7.5 Combinations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 280 18.8 Application to Portuguese . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 282 18.9 Application to Danish . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 283

18.9.1 Type . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 283 18.10Application to Greek . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 284

18.10.1 Type . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 284 18.10.2 Gender, Number and Case . . . . . . . . . . . . . . . . . . . . . . . . . . . 284 18.10.3 Function . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 285 18.10.4 Form . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 285 18.11Application to Swedish . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 286

18.11.1 Type . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 286 18.12Application to Catalan . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 287 18.13Application to Irish . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 288

18.13.1 Type . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 288

19 Interjections 289

19.1 Application to Italian . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 289 19.2 Application to German . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 289 19.3 Application to English . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 289 19.4 Application to Dutch . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 289 19.5 Application to French (Corpus) . . . . . . . . . . . . . . . . . . . . . . . . . . . . 289 19.6 Application to French (Lexicon) . . . . . . . . . . . . . . . . . . . . . . . . . . . . 289 19.7 Application to Portuguese . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 289

Recommendations 15 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R

19.8 Application to Danish . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 289 19.9 Application to Greek . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 289 19.10Application to Swedish . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 289 19.11Application to Catalan . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 290 19.12Application to Irish . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 290

20 Unique membership class 291

20.1 Comments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 291 20.2 Application to German . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 291 20.3 Application to English . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 291

20.3.1 Particle-Type . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 291 20.4 Application to Dutch . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 291 20.5 Application to French (Corpus and Lexicon) . . . . . . . . . . . . . . . . . . . . . 291 20.6 Application to Greek . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 292

20.6.1 Particle-Type . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 292 20.7 Application to Irish . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 292

21 Residual 293

21.1 Comments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 294

21.1.1 Type . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 294 21.2 Application to Italian . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 295

21.2.1 Type . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 295 21.3 Application to German . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 295

21.3.1 Type . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 295 21.4 Application to English . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 295

21.4.1 Type . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 295 21.4.2 Number . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 295 21.5 Application to Dutch . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 295 21.6 Application to French (Corpus) . . . . . . . . . . . . . . . . . . . . . . . . . . . . 296 21.7 Application to French (Lexicon) . . . . . . . . . . . . . . . . . . . . . . . . . . . . 296 21.8 Application to Greek . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 296

21.8.1 Foreign Word . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 296 21.8.2 Type . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 297 21.9 Application to Catalan . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 297 21.10Application to Irish . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 297

21.10.1 Type . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 297

22 References 298

Recommendations 16 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R 1 Authors

Monica Monachini Nicoletta Calzolari (Coords.) Istituto di Linguistica Computazionale - CNR Via della Faggiola 32 Pisa Tel: +39.50.56.04.81 Italy Fax: +39.50.58.90.55 56100 E-mail: glottolo@ilc.pi.cnr.it

2 Other Contributors Joan Soler (CA) Anna Braasch and Ole Norling-Christensen (DA) Tilly Dutilh-Ruitenberg (DU) Geoffrey Leech and Andrew Wilson (EN) Jean-Marc Lang'e, and Jean Veronis, Liliane Khouri and Christine Meunier (FR) Anne Schiller (GE) Penny Labropoulou and Maria Gavrilidou (GR) Eoghan Mac Aog'ain (IR) Paula Guerreiro (PO) Nuria Bel and Marta Villegas (SP) Pernilla Danielsson and Jerker J"arborg (SW)

3 Commentators Susan Armstrong, Gerardo Arrarte, F. Bacelar de Nascimento Lynne Cahill, K.A.C. Depuydt, Pieter de Haan, Gerald Gazdar, Ramia Hatzidaki, Ulrich Heid, Dirk Heylen, Philip King, Lothar Lemnitzer, Aurora Mart'in, Wolfgang Paprottee, Graham Russel, Helena Soares, Petra Steiner, Simone Teufel, Dan Tufis, Ursula von Rekowsky, David Wigg, Annie Zaenen

4 Introduction 4.1 Standards and Shareable Resources: the Area of Morphosyntax Computational lexicons represent an area where a pressing need is felt for the provision of basic large-scale resources which must be actually reusable, interchangeable, shareable and also integrable in a multilingual environment1.

1The present introduction has been revised at the beginning of 1996. Hence, the feedback received after the first cyle of comments promoted by EAGLES - during which suggestions on the preliminary EAGLES recommendations have been sent to the EAGLES editors by experts in the field - and the experience gained within the EAGLES validation phase and the MLAP-PAROLE project have been taken into account.

Recommendations 17 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R In terms of costs and effort it is quite expensive to produce such resources in detail. In the construction of large-scale resources and in the definition of annotation strategies, reusability is an essential requisite to pursue (Calzolari 1994).

If we want to produce and exchange Natural Language resources which are effectively shareable, and possibly usable by the same set of basic tools, with results of processing which are really comparable and integrable, we should aim at designing and producing these resources according to a common set of consensual and broadly agreed specifications, i.e. standards (Calzolari and Zampolli 1994).

As discussed in Calzolari and Monachini 1995, the Morphosyntax area was among the first in which EAGLES concentrated its efforts, because it is a more mature area, where a lot of work has already been done, many systems, approaches, and data exist for many languages: this constitutes a solid platform for making reasonable and acceptable proposals for standards. The work previously undertaken in the NERC project (Monachini and Oestling 1992a and 1992b, NERC 1994), proved that this field is an area that permits the agreement on common specifications. The tagsets already existing for many languages and the encoding practices for many computational lexicons constitute the necessary basis of accumulated experience.

Moreover, morphosyntactic annotation constitutes somehow the basic level of linguistic description and could be considered a prerequisite for further and more complex kinds of analysis. Almost all of the systems and applications do require this level of linguistic description. In this area, hence, standard conventions are welcomed by quite a large number of users in all sectors of LE, but also in the literary and humanities field.

The objective of the present document is to propose a common core set of morphosyntactic distinctions to be encoded in lexicons of the European languages.

5 Methodolgogy Adopted 5.1 Interdependence between the Lexicon and the Corpus View The awareness of the interrelation between morphosyntactic encoding in the lexicon and morphosyntactic annotation in corpora, i.e. between lexicon specifications and corpus tags (see below 3.1 for further details), guided the approach adopted in the two working groups2. The two EAGLES subgroups on Morphosyntax, in the Lexicon and the Corpus Working Groups, have always worked in close cooperation, and much attention was paid to the definition of compatible sets of attributes and values. Two documents have been carried out by the two subgroups - one on lexicon specifications (the present report) and the other on corpus tagsets (Leech and Wilson 1994). The recommendations must be seen as two complementary sets, examining similar sets of phenomena from different viewpoints, a more general one (the lexicon) capable of being directly mapped onto an application oriented one (the corpus).

Special thanks are due to Mr. A. Wilson, who revised the Introduction as far as English is concerned.

2The two EAGLES subgroups working on Morphosyntax in Corpora and in Lexicons agreed on the approach

towards morphosyntactcic specifications presented in this document and in the Corpus document at a Joint Meeting of July 9th and 10th 1993 in Pisa.

Recommendations 18 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R 5.2 Bottom-up approach In agreement with the general EAGLES methodology, a bottom-up approach, taking as starting point existing practices in the main lexical and textual projects world-wide (both in lexicon specifications and in corpus tagsets) has been adopted.

The procedure followed was:

1. to survey the main encoding practices for morphosyntactic description in lexicons and in

corpora (section 2.3) with the aim of deriving a consensual proposal from their comparison (2.4);

2. to carry out a testing of the proposal, by applying it concretely to the European languages

(2.5 and 4).

This allowed us to highlight the areas of common ground and some aspects of discrepancy between the different systems for classifying morphosyntactic phenomena (2.3.1), and to provide a common nucleus of morphosyntactic distinctions. These are proposed in a feature-based notation, in the form of attribute and value pairs (2.4.1).

The present set of recommendations was tested with respect to all EC languages. Moreover, the specifications were adopted - among others - within the two large EC projects LRE-MULTEXT (2.6.1) and MLAP-PAROLE (2.6.2), and - as a validation task - have been translated into formal and explicit guidelines for morphosyntactic encoding in lexicons for four European Languages - English, French, German and Italian (2.7).

5.3 The Survey and Comparison Phase The morphosyntactic descriptions and encoding schemes involved in the comparison phase are the following:

ffl on the side of lexicons:

(a) the MULTILEX model, - 1st row in the tables (2.4.1) - as presented in the the

Final Report on Morphology (MULTILEX 1993),

(b) the GENELEX model - 2nd row in the tables - for the encoding of the morphological and syntactic levels in a lexicon, originally studied for French (GENELEX Apr.1993, GENELEX Sept.1993),

(c) the specific AlethDic application of the GENELEX model (GSI-Erli 1993) - 3rd

row in the tables;

ffl on the side of corpora:

(a) the proposal of a consensual nucleus of morphosyntactic information encoded by

the most common existing tagging practices - 4th row in the tables - presented in the framework of the NERC Project (Monachini and Oestling 1992a and 1992b), which took into account the following tagsets: Penn Gothenburg and Brown for American English, BNC, LOB, Lancaster and ENGTWOL lexicon for British English, ILC-DMI and EUROTRA for Italian, INaLF for French, Uit den Boogaart for Dutch;

Recommendations 19 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R

(b) the scheme proposed by Leech and Wilson - 5th row in the tables - as an outcome

of a joint meeting of the Lexicon and Corpus Groups in Pisa (Leech and Wilson 1993). Note that this proposal is a step forward with respect to that of NERC since, besides corpora, it also takes into account the first requirements for lexicons which emerged from the discussion at the joint meeting in Pisa.

Additionally, it should be noted that the NERC proposal also took into account the list of common morphological features proposed within the TEI by the Linguistic Analysis Committee (TEI AI1W2 1991).

The comparison is displayed, category by category, by means of synoptical tables (6.1, 6.2, 6.3 etc.) subdivided into two zones - the first one devoted to the comparison (see 2.3.1) and the second one to the EAGLES proposal (see 2.4.1) - containing the relevant features, followed by some explanatory notes and comments.

5.3.1 Description of the first zone of the tables The tables representing a category are structured as follows:

ffl the 1st vertical column on the left contains the names of the encoding systems under

analysis;

ffl the top horizontal row displays the morphosyntactic category (PoS) considered (first item

on the left), and the relevant morphosyntactic information, presented as attribute names;

ffl each column has the name of an attribute and lists the relevant values that are used

within each system. If the cell is left empty, this means that the system does not mark the information.

PoS Attribute Attribute Attribute Attribute ... MULTILEX value value value value ...

GENELEX value value value ...

AlethDic value value value ...

NERC value value value ... EAGLeech value value value ...

EAG-L0 obligatory: PoS EAG-L1 recommended: minimal common core set of features EAG-L2a optional: info common to languages,

either not usually encoded or not purely morphosyntactic EAG-L2b language-specific: language-specific info

Recommendations 20 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R 5.4 The Proposal Phase The common set of specifications, resulting from an analysis and evaluation of the results of the comparison, and constituting the EAGLES proposal is contained in the bottom box, a set of rows named EAG-L0, EAG-L1, EAG-L2a and EAG-L2b in the second zone of the tables. The features are indeed articulated on different levels, corresponding to different degrees of "obligatoriness". As already pointed out (see Leech 1992, Monachini and Oestling 1992b, Leech and Wilson 1993b), different levels of constraints can be isolated in the morphosyntactic encoding of a category and, therefore, different levels of standardization can be suggested.

5.4.1 Description of the second zone of the tables In the EAGLES proposal, the specifications appear as follow:

ffl Level 0 (L0) contains only the category (or PoS), i.e. Noun, Verb, Adj. etc., presented

also in the document by Leech and Wilson (1993b) as the only "obligatory features";

ffl Level 1 (L1) presents the grammatical features, the agreement features such as Gender,

Number, Person, etc., which are usually encoded in lexicons and corpora: these are considered as "recommended features" (in Leech and Wilson defined as "optional") constituting the "minimal common core set of features" for the PoS;

ffl Leech and Wilson "Special Extensions" are presented here on two levels; they are in fact

exemplary of current practices from two different points of view:

- Level 2a (L2a) contains information which is pertinent to all or many languages and

is considered useful and easy to standardize, but is either not yet usually encoded by current practices or not purely morphosyntactic (e.g. countability for nouns): these are to be considered as "optional features". As a rule of thumb, a value is put here if it is relevant to more than three languages.

- Level 2b (L2b) presents "language-specific features", not belonging to the set of

recognized common features. The values presented in this row are labelled with the initials of the language they are pertinent to.

PoS Attribute Attribute Attribute Attribute ... MULTILEX value value value value ...

GENELEX value value value ...

AlethDic value value value ...

NERC value value value ... EAGLeech value value value ...

EAG-L0 obligatory: PoS EAG-L1 recommended: minimal common core set of features EAG-L2a optional: info common to languages,

either not usually encoded or not purely morphosyntactic EAG-L2b language-specific: language-specific info

Recommendations 21 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R The names of the attributes and values are presented in the tables in an abbreviated form, due to the limits allowed by the page: almost often, the abbreviations are self-explanatory (e.g. Gend stands for Gender and Numb for Number). Anyway, the full names of the attributes with the relevant values are listed in the "Comment" section which follows each of the tables. The abbreviations of those attributes that do not appear clear from the tables can be, hence, desumed from there.

As they are intended now, the abbreviations do not represent in itself a proposal of how to encode concretely these attributes in lexicons, i.e. they are only labels to represent a distinction that in the various European concrete practices can be marked using different simbols: e.g. for the value of Number singular, singular, sing, sg, s, 1 can be used. What is important is the mappability from a concrete system of labels to the set of labels proposed here. This can serve, therefore, as an interface or interchange proposal.

Having a multi-layered or hierarchical structure, instead of a flat one, gives more flexibility to the proposal and allows the user to choose the most appropriate level of encoding (which may vary, e.g. according to different applications). The idea is that, going from Level 0 to Level 2 the amount of information is increased and more granularity is achieved. The use of all three levels allows the representation of more information, whereas a description at Level 1 is a subset of a description resulting from the application of all three levels. Moreover, a multi-level proposal provides an easy framework for specific extensions and updating, as well as permitting a comparison of different schemes (at the lowest common level).

5.5 The Application Phase For each category, an application of the common proposal (which corresponds, as far as possible, to the union of the analysed schemes) to a large number of European languages has been carried out. Examples from the specific language are given for each attribute value, and, when necessary, criteria for the application of the feature are described.

5.5.1 Description of the Tables used in the Applications Category by category, the specifications proposed are tested on real corpora and lexicons in a large number of European languages.

In these language-specific sections (6.1, 7.1 etc.), each morphosyntactic feature pertinent to a category in a specific language is represented in the form of a table: the attribute name is presented in bold characters, the values are listed in italics and examples are provided, together with the information on the corresponding tag in a language-specific lexicon encoding scheme or corpus tagset (if available).

ffl If a language does not have some values for a given attribute, the cell of the example (and

of the tag) is left empty.

ffl If the attribute has, in a particular language, more values than in the proposed common

system, the extra-value is inserted in the second part of the table (with the note l-spec to

Recommendations 22 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R

the left). ffl If, in a particular language, some features are not at all applicable, this will be dealt with

in the "Comments" section; the feature will not be represented as a table.

We give here an example of such a table, i.e. the feature "Number" for Nouns in the Italian section:

Attribute value It. example It. tag

Number singular (il) libro S/ms

plural (i) libri S/mp

l-spec invariant (la/le) attivita' S/fn

For the semantics of the value "invariant", the reader is referred to the discussion presented in the section 6.1.

5.6 Interaction with some European Projects As already mentioned in the introductory part, the experience gained by the application of the EAGLES proposal in some European and national projects (see section 5 for an exhaustive list) had a strong influence on the set of the recommendations. The specifications, hence, had been cyclically revised according to the requirements emerged while working in the framework of these projects.

Below, the experience of MULTEXT and PAROLE are summarized and the reciprocal influence between these two projects and EAGLES is reported. Mention to the work carried out in the validation phase of the Lexicon WG is also made. It is worth recalling here that very useful feedback has been received and the impact of the present proposal has been measured also from its use in other projects.

5.6.1 The LRE-MULTEXT Experience The main objectives of the MULTEXT project (MULTEXT Tech.Ann. 1993) were the definition and the implementation of a set of tools for Corpus-based research and applications, and the production of a corpus in a multilingual framework. Tools and resources were developed on the basis of operational standards and in the light of the conventions which are being defined by the major international projects dealing with the issue of standardization.

One of the MULTEXT tasks dealing with annotation conventions and hence strongly connected with the work presented in this document, aimed at formulating: (i) common specifications and a common notation for the MULTEXT lexicon, and (ii) a tagset for the MULTEXT corpus on which the tools will run.

Recommendations 23 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R The MULTEXT partners involved in this task3 had carefully evaluated the proposal which was defined within EAGLES - the proposal presented in a preliminary version of the present document - for each PoS at Level-1 (the level of recommended features). After a global evaluation of the EAGLES proposal, also taking into account the different grammatical traditions and the different language requirements, they checked if the features suited the description of their respective languages and added those features and/or values needed at the language specific level. All the partners performed the evaluation by translating their existing - or still under development - lexicons into the EAGLES features and values, i.e. applying the proposal concretely to their languages and providing examples of all the admitted combinations of values for each category. In such a way, constraints on the application of the values emerged (see the application to the French Lexicon 4.6 and 6.8, 7.8, 8.8, etc., included in the present report).

The MULTEXT experience turned out to be a very important test-bed for the EAGLES Lexicon proposal: - a large core of lexicon specifications proved to fit the description of all the six MULTEXT languages (Dutch, German, English, French, Italian and Spanish); - the cycle of testing and concrete application stressed the need of further specifications at the language-specific level.

An essential change affected the class of Pronouns, which in a preceding version of the EAGLES Lexicon document incorporated the Determiners; the previous merging of the two categories (at least at L1, with the possibility of splitting the two categories at a more fine-grained level) seemed, in the first instance, to be the best solution to cope with the requirements of many corpus practices - that keep the two categories undistinguished - and attempted to reconcile lexicon specifications and corpus tagsets. This choice, however, was rightly felt to be too corpus-oriented and the MULTEXT partners expressed their opinion in favour of having, at the lexicon level, two different categories for Pronouns and Determiners. Lexical descriptions should aim, indeed, at a general and fine-grained description of the language which is independent from particular applications, while, given a set of practical reasons - state-of-the-art tagging techniques and computability (see Bel Calzolari and Monachini, Coords., 1995) -, broader categories are to be preferred for the tagsets where "collapsing" decisions are to be made.

The requirements that emerged from the MULTEXT feedback, therefore, are reflected in the the lexicon specifications of the present version of this document.

It is worth also mentioning here that the EAGLES/MULTEXT specifications were extended and adapted to cover the requirements and peculiarities of a number of Eastern languages, within the framework of the COPERNICUS-MULTEXT-East Project. In particular, the common core specifications proved to be suitable to the Eastern languages and the phylososphy of the multi-layered proposal effective for extensions and adaptations (Monachini ed., 1995b).

3Istituto di Linguistica Computazionale - Pisa, Italy (Coord.); Laboratoire Parole et Languge CNRS - Aix-enProvence, France; Fundacion Bosch Gimpera - Barcelona, Spain; Institut Dalle Molle pour les Etudes Semantiques et Cognitives - Geneva, Switzerland; Universitaet Muenster - Muenster, Germany; Siemens - Barcelona, Spain; Stichting Taaltechnology - Utrecht, The Netherlands.

Recommendations 24 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R 5.6.2 The MLAP-PAROLE Experience Within the MLAP-PAROLE project, the EAGLES recommendations for Morphosyntactic encoding in Lexicons were taken as the starting point for finding and establishing the framework in which either to reuse existing lexical data or to start creating new ones, according to harmonized criteria, ensuring their reusability.

The definition of the minimal initial technical conditions for starting the activity of actually providing Written Language Resources was achieved by translating the EAGLES recommendations into operational standards (Monachini ed., 1996). The first action of the PAROLE partners towards the definition of operational standards was the production of language specific instantiations of the EAGLES specifications, having in mind the practical criteria and following the steps listed below.

- In the multilayered EAGLES proposal, the most appropriate level of encoding for lexicon resources was recognized, i.e. the the set of values (i) pertinent to each language, (ii) feasible to encode within the constraints of time, money, and lexicon resources already available, and (iii) relevant for a large range of NLP applications was isolated.

- Once the attribute-value pertinent set was circumscribed, the logic relationships between the values of the system pertinent to the language in analysis were formalized, in the cases in which the values belong to different levels.

- Furthermore, the dependencies and the constraints in the application of an attribute or a value in presence of a given attribute or value were also specified.

These last two points are, in our opinion, particularly crucial. Within EAGLES, all the proposed lexicon specifications have been collected together in a large repository, articulated on different levels of obligatoriness (see section 3.4 for further discussion on this point). Each language-specific application shows the pertinent set to one language and permits us to deduce the relations between values on the basis of the semantics and the actual exemplifications. This, however, can cause misinterpretation. In the PAROLE task, we gone a step further and, for each language, we provided a formalized system readily interpretable with the aim to obtain a structured inventory: the logic relationships between the values have been specified in order to make others able of consistently interpret their semantics itself.

In order to highlight the difference with respect to the EAGLES language-specific applications, the concrete translations of the EAGLES specifications into operational standards carried out by the PAROLE partners on their languages according to the criteria listed above have been called "instantiations".

5.7 The Guidelines of the Validation Phase In the validation phase of the Lexicon Morphosyntax work within EAGLES, the production of guidelines, i.e. clear tests and explicit criteria for the encoding of lexicons on the basis of the EAGLES recommendations, has been undertaken for English, French, German and Italian. The Guidelines are available now for French (Von Rekovsky U. "ELM-FR: A typed French incarRecommendations 25 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R nation of the EAGLES-TS Definition of Lexicon Specifications and Classification guidelines"), German (Teufel S. "ELM-DE: A typed German incarnation of the EAGLES-TS Definition of Lexicon Specifications and Classification guidelines"), English (Teufel S. "ELM-EN: EAGLES Validation (WP4) Task on Tagset Mapping") and Italian (Monachini M. "ELM-IT: A typed Italian incarnation of the EAGLES-TS Definition of Lexicon Specifications and Classification guidelines").

In these documents, together with the guidelines for the encoding of the specifications of the EAGLES recommendations, an important aspect in the production of practical criteria for the encoding of lexicons is also highlighted: the dependencies between different attribute-value sets and the constraints in the application of an attribute/value in presence of another attribute/value. These phenomena are represented in terms of typed hierachies of each feature. Why a typed hierarchy has been chosen? That has been considered a good means to translate those dependencies and constraints into a formalism - which makes use of Type Feature Structures (TFS) -, and to handle them with rigorous computational techniques rather than leaving them to human control. In the creation of these typed hierarchies some values/attributes proved to be necessary for representing the structure of the attribute-value and handling computationally dependencies: these features have been integrated in the tables and are now part of the recommendations.

6 Some relevant aspects for lexicon specifications We deal in this section with a number of aspects that are of relevance as introductory general remarks with respect to the goal of providing a common core of lexicon specifications.

The issues that we want briefly to discuss concern aspects such as the relationship between lexicon specifications and corpus tagsets, the concept of "common category", the objects that are described and the level of description, the descriptive approach (general vs. language specific) and its implications for problems like monotonicity, redundancy, etc., the introduction of constraints at the language-specific level, and so on.

6.1 Lexical specifications vs. corpus tagsets The interdependence between lexicon and corpus is a very important aspect for any action aiming at creating lexicons and/or tagsets to be shared and made available to the community. The background motivation to this was essentially the view of corpus tagging as just one of the possible applications of a Computational Lexicon, which has to be seen in a more neutral context as an application-independent set of lexicon specifications. Corpus tagging is in fact the first obvious application of a Computational Lexicon and cannot be developed on an independent basis: both the lexicon specialists and the corpus specialists feel that it is very important to reconcile the two views.

The difference in perspective betwen the lexicon specification area and the corpus annotation can be reflected at the level of terminology: - the term "features" is preferred when talking about lexicon descriptions; - the terms "tag" and "tagset" are used for the information associated with words in context,

Recommendations 26 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R i.e. in corpus annotation. For the sake of re-usability, the lexical descriptions should be (as far as possible) independent from specific applications, and should aim at a general description of each language. The actual corpus tags depend on at least the following: (1) the lexicon features, and (2) the capabilities of the state-of-the-art taggers to disambiguate between different lexicon descriptions or different types of typical homographies present in different language types.

Therefore, morphosyntax can be encoded in a lexicon with fine granularity, while a set of corpus tags usually reflects broader categories.

The corpus tags are, in fact, developed for each language with a particular application in mind, that of producing a corpus tagged for part of speech (and possibly other morphosyntactic information) by means of automatic disambiguation. It would be ideal to tag a corpus with the lexical descriptions themselves. However, it is well known that this is considerably beyond the capabilities of state-of-the art tagging techniques. Corpus tags are, therefore, to be seen as a kind of underspecified lexical tags. There are two reasons why we may want (or need) to underspecify corpus tags:

1. Experience shows that some distinctions are difficult to get right with a high rate of accuracy. (For example, in some languages, the disambiguation of indicative present and subjunctive present in a corpus is extremely difficult by automatic means).

2. In order to train the tagger, we need statistical tables (based on co-occurrences of tags). If we have a large tagset, we need a very large corpus to train the disambiguator, in order to observe rare co-occurrences. For example, in the proposal for French presented in the MULTEXT document (Bel, Calzolari and Monachini eds. 1995), there are 249 different lexical descriptions, but only 74 collapsed corpus tags. Experience (Church, Penn Treebank, IBM France, etc.) shows that the tagset should be under 100 tags. In fact the Penn Treebank collapsed many tags compared to the original Brown, and got better results.

Two other observations are of relevance as regards the relation between lexicon specifications and corpus tags.

(a) Sometimes tagging classes are in reality different from lexical descriptions. For example, classes for punctuation are needed and certain types of semantic or pragmatic or lexical information can be present in the tags (e.g. the days of the week).

(b) Furthermore, the "collapsing" decisions are often language dependent and therefore it may not be appropriate to have completely identical tagsets across languages. We must preserve certain language-specific peculiarities (e.g. if certain distinctions can be easily maintained by an automatic tagger, it may be useful to preserve them).

Recommendations 27 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R 6.2 The concept of "common" category A debate between the MULTEXT partners generated some possible definitions of "common category" on which is worth reflecting: - the same meaning of a category in different languages implies that it enters in the same combinations (distributional definition of common categories); - a category is common "if it yields isomorphic partitions of word-lists or lexicons under translation (module exceptions)". A category system is common to languages L0 and L1 if members of class C in L0 translate as members of C in L1 (translational definition of common categories); - a common category among languages is "a category which conveys the same linguistic information (i.e. stands for the same linguistic phenomenon) in all the languages in which it is used"; - "a word class is common across the languages if it reflects a linguistic category/phenomenon which is either morphologically or lexically expressed in at least two of these languages, even though there may not be a one-to-one lexical relationship for this in translation".

We could define common categories as "those whose members satisfy the same criteria and tests": this crucially implies a clear definition of criteria for the recognition of their members (see the Guidelines produced in the Validation Phase 2.7). In absence of such explicit criteria, we can empirically recognize "common" categories which are relevant in the morphosyntactic descriptions of a number of European languages. These common categories are usable and actually used in the largest lexicons and corpora and have in general the same "meaning" in the different languages, even though the property of "commonality" holds more for open classes and poses more problems for function words or closed classes.

Taking in mind this simple equation of "common categories" and "actually used categories", their "adequacy" in terms of user requirements can also be achieved: it is "empirically" obtained through the bottom-up process of looking at the largest and most "used" lexicons and tagging schemata (as was done here). Behind these lexicons and annotation schemata there have been many different types of users (of lexicons and of annotated corpora).

In fact, with the above caveat in mind, it was found from the analysis of the schemes that a lot of "commonalities" proposed here are relevant for many languages. Different schools and traditions of languages can agree on such a simple set of features.

6.3 Objects described and level of description The typical objects that are described in this proposal are lemmas (even though we do not deal here with lexical decisions as to what is to be to considered as a lemma) and word-forms at the morphosyntactic level. This essentially includes information on the grammatical category or part of speech, their subtypes as found in lexicons, and inflectional phenomena to be encoded in attributes such as gender, number, tense, etc.

What we propose here is the basic set of core features, derived from a detailed analysis of the major European lexicon and corpus projects; we do not aim at giving a completely worked out set of specifications ready to be implemented as such. This task is to be left at the level of the

Recommendations 28 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R language specific development of concrete application lexicons. One problem which we encounter as far as objects described are concerned, is how to deal with the two complementary phenomena of a) grammatical categories split into more than one graphical unit (multiwords or discontinuous words), and b) graphical units composed of more than one grammatical category (e.g. contractions). Examples of the first type are found in section 3, while examples of the second type are found in section 8. In general EAGLES recommends handling multiwords as belonging to a single grammatical category, and contractions as two separate grammatical categories, but the option of a different treatment is left open.

6.4 Descriptive approach over different languages wrt monotonicity The general approach underlying the EAGLES proposal follows the ET-7 (Heid and McNaught Eds. 1991) proposal of looking for the basic phenomena at each level of linguistic description, going to the more granular level, and providing the more detailed set of features able to encode the relevant phenomena.

This approach is taken here over the set of European languages, trying to reach the same level of granularity for the description of each of them. For each language, the most common practices for lexicons and corpora were considered.

The obvious consequence of the two approaches together (i.e. granularity and many languages) is that a large repository of potentially useful lexicon specifications is formed where all the features which are necessary for the description of the basic phenomena in the many languages considered are juxtaposed to each other. Given the differences in the different language-specific systems, where each system has its own set of constraints, the large collections of features - summarized in the Tables at the beginning of the description of each grammatical category - do not, and cannot, constitute a consistent system to be implemented as such, but are a redundant inventory of all the possible features relevant for that category across the different languages (an ET-7 conformant big repository, according to the "data pool" model). Each language-specific system can afterwards be implemented as a specific application of the general redundant set, by picking up the features and values appropriate for its system.

At the general level of the large synoptical tables there is, by definition (i.e. by the very way in which they are constructed), no property of monotonicity (and no necessity of it), but, in contrast, there is redundancy and conflicting values may also be found there.

It is only at the level of the language-specific instantiations - considered as applications where the problem of the representation formalism also arises - that monotonicity can be looked for. Applications of the general tables are given for most of the European languages as a reinforcement of the EAGLES proposal, and here - in particular when these will be detailed for real lexicon building - the different constraints have to be specified, as far as possible, between different features, the range of pertinent values for each feature have to be made explicit and aspects of the hierarchical organisation of the features have to be solved (see the Guidelines produced within the Validation Phase, 2.7. and the PAROLE contribution, 2.6.2).

Recommendations 29 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R Within this approach it is assumed that not all the values presented in the general Tables are relevant for "all" the languages. There will be cases in which some do not apply, and this has to be made clear in the language-specific applications.

7 Language-specific Applications A few general information about the authors of the language-specific applications and the lexicons and/or corpora which the proposal has been applied to, are provided below. More details - when available - are given in the relevant language-specific sections following each category.

7.1 Application to Italian The application to Italian has been carried out by M. Monachini and N. Calzolari (Monachini and Calzolari 1994).

When speaking of the Italian lexicon, we refer here to the Italian Machine Dictionary - DMI (Calzolari et al. 1983, Monachini 1992); when speaking of the Italian Corpus, we refer to the Italian Reference Corpus (Bindi et al. 1991). The column "It.tag" in the tables displays the actual tag used in the Italian Machine Dictionary and in the Corpus tagset to represent the attribute and the value.

7.2 Application to German Sections on German have been written by A. Schiller (Schiller 1993).

The application to German takes into consideration the tagset developed in the project Text Corpora und Erschliessungswerkzeuge - TC and at the IMS Stuttgart (hereafter called the "IMS-Tagset") in collaboration with the project korpusunterst"utzte Entwicklung lexikalischer Wissensbasen f"ur die Computerlinguistik - ELWIS at the University of T"ubingen.

Recommendations 30 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R 7.3 Application to English Sections on English have been contributed by G. Leech and A. Wilson (Leech and Wilson 1994b).

The English tagset suggested here does not follow, except in a very broad sense, the practices now established in the tagging of English corpora. There are existing tagging systems - for example, the CLAWS5 and CLAWS7 tagsets employed for the British National Corpus - which might have been used as a model. However, the difficulty is that the morphosyntax of English is exceptionally simple, and many of the distinctions often represented in English tagsets are not strictly morphosyntactic: for example, the distinction between attributive and predicative adjectives. On the other hand, it is usual practice to assign to the base form of the English verb just one tag, or at most two, rather than to represent all the values of number, tense, person, and mood which would be required by most direct adherence to the EAGLES guidelines. In practice, therefore, the English tagset presented here is something of a compromise solution which will enable English morphosyntactic information to be integrated in the document.

7.4 Application to Dutch The application to Dutch has been carried out by T. Dutilh Ruitenberg (Dutilh Ruitenberg 1994)4.

Most of the tags described in the sections devoted to Dutch application are selected from two different CELEX lexical databases: some of them are taken from the Syntactical information columns of the Lemmas Lexicon and some from the Inflectional information columns of the Wordforms Lexicon. To understand the term column it is necessary to understand that the databases consist of rows divided into dozens of columns, each column containing a value tag. Each row contains in its columns the information belonging to one Lemma (Lemmas lexicon) or to one word form (Wordforms Lexicon). The Syntactical information in the Lemmas lexicon is represented by Yes/No tags or by abbreviated name tags. As well as such Yes/No tags, the Wordforms lexicon presents a different kind of tags, called Flection Type tags. These are single or composed tags, each element representing one aspect of the inflected wordform.

The examples given in the CELEX tables of the Dutch application are deliberately taken from the DUTCH LINGUISTIC GUIDE (Burnage 1990). If there were no examples the authors made them up herself. This is always marked.

The CELEX system allows the attribution to one word form of double or triple tags, separated by slashes, such as: `geo./pers.' and `hebben/zijn' and `intrans./trans./wederk.'.

It should be noted that there are two representations of each syntactic tag in the Lemmas lexicon: a numeric one and an alphanumeric one. In the tables of the application the alphanumeric tags are only found.

Some categories are not present in the CELEX databases mentioned above. To fill this gap, some distinctions labelled in the applications as `Proposals for Dutch', either based on the German Tables, or on the MULTILEX table, and always based on Dutch traditional grammar are

4Special thanks are due to J. v.d. Voort v.d. Kleij for substantial comments.

Recommendations 31 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R presented. There is another CELEX database, containing still more detailed information on subcategorization and subclassification of a syntactic as well as a semantic nature. This database is not yet fully described, but the author did use this source for some tables in the application.

The CELEX tags are in the Dutch language, whereas the tags in the ten `Proposals' are Latinderived or English.

For Dutch, a report has also been contributed by Dirk Heylen (Heylen 1994) from Utrecht. In his document Heylen makes useful comments mainly concerning the application to Dutch, but some general considerations and questions are raised as well. (This feedback is taken into account and incorporated in the Dutch section by the Dutch correspondent.)

7.5 Application to Spanish Sections on Spanish have been written by N. Bel and M. Villegas (Bel and Villegas 1993).

The Spanish EUROTRA dictionary and the Spanish tagging used in Eurotra's dictionaries are taken as reference point.

7.6 Application to French The EAGLES proposal has been analysed from a corpus point of view by Jean-Marc Lang'e (Lang'e 1994) of IBM France, and its applicability to a French Lexicon has been tested by the Veronis group of CNRS (Veronis et al. 1994a and 1994b). Their reports also constitute a first proposal for encoding a lexicon to be used as a basis for the specific application of corpus tagging within the MULTEXT project.

The sections on corpora describe the application of the EAGLES proposed model to a French corpus, taking as a reference point the tagset developed and used at IBM France Scientific Center, in particular by the speech group. For ease of use, this tagset will be referred to as the IBMF tagset. Despite this focus on a particular tagset, general problems will also be mentioned, even when they do not lead to any discussion of the French language (for example, everybody will agree that nouns do not bear case information in French). Tagsets such as the IBMF tagset have a certain bias since they are used for the very specific purpose of predicting the exact part-of-speech of words in a corpus; in other words, they are used for modelling the language at the morphological level, whereas a lexicon tagset would be developed for describing the language. For example, this tagset doesn't cover the full set of features for verbs (mood, tense, etc.), for two reasons:

ffl the graphic form of the verb will help -if necessary- determine the "missing" features ffl the addition of tags with these specific features would not improve the language model's

essential capability: that of correctly predicting other tags.

For similar reasons, several corpus-oriented tagsets for the same language might differ considerably, depending on the goal pursued (e.g. speech recognition, terminology identification,

Recommendations 32 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R etc.) or on the type of language modelling used (e.g. stochastic vs. rule-based models, etc.). So the reader should be careful and consider this contribution for French only as one possible application, not as an attempt to describe a universal solution.

The features are listed in the following order: 1) first the different EAGLES attributes/values applicable to the IBMF tagset, 2) then the EAGLES features that are NOT applicable to the tagset or to French, 3) finally the specific items for which there is no attribute/value in EAGLES (those which are not language-specific but rather tagset-specific and therefore do not need a new EAGLES attribute).

When only a part of the tag name is relevant to the specific attribute-value pair considered, this part will be put in bold font in the Tag column (e.g. in tag SUBSFS, SUBSFS applies to substantive, SUBSFS to feminine and SUBSFS to singular).

7.7 Application to Portuguese The sections on Portuguese have been written by P.Guerreiro (ILTEC) (Guerreiro 1994).

The author reports on a skeleton evaluation of the EAGLES proposal with respect to the Portuguese language. A comparative approach was adopted. As a main basis, the specifications used in the morphological layer developed by ILTEC in the framework of the GENELEX project (Guerreiro (ed.), 1994), instantiated in a demo lexicon of about 5000 entries, were taken into account and compared with those provided by EAGLES. Ongoing work on a morphological analyser for the Portuguese language was also sometimes taken into account.

7.8 Application to Danish The application to Danish has been contributed by Anna Braasch of CST, Copenhagen (Braasch 1994); the description of the Danish Word Bank has been contributed by Ole Norling-Christensen.

In all the tables in the Danish sections the value names are used in accordance with the Eurotra Dictionary Encoding Manual for Danish (EDEMD) and/or with the appropriate namings used in traditional dictionaries. The names are also harmonised with the tagset planned for the corpus tagging tool for Danish.

7.9 Application to Greek The application to Greek has been carried out by P. Labropoulou, M. Gavrilidou and E. Mantzari, based on (Aglamissis et al. 1994) and (Mantzari & Gavrilidou 1995).

The Greek Morphological Lexicon described in this report is the ILSP Morphological Lexicon which is used for the purposes of morphological annotation of the Greek Reference Corpus. At present, this lexicon is under review so that it can be used for other applications, as well.

Recommendations 33 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R The entries of the lexicon are classified on the basis of the grammatical category and the information included in it constitutes of:

- for inflected forms: stem, code number for inflectional paradigm, place of stress (number of stressed syllable), optional diaeresis,

- for uninflected forms: word form, number of stressed syllable. Given the above classification, information on the morphological features of each entry is derivable from the inflectional classes (e.g. gender, number and case for nouns). Further types of information depending on the grammatical category of the entries (e.g. type of prepositions, common/proper noun, etc.) are explicitly coded for each entry. More details on the information is presented in the following sections, separately for each category.

Therefore, in the tables that are presented for each entry, the column presenting the tags shows both the tag that is generated by a special mechanism for each word form of the lexicon, and the tag that would accompany this word if found in a text.

Greek examples are transcribed in the Latin alphabet, using the system adopted for Eurotra. This system for transcription adopted the optical similarity of the graphemes used by the Greek language. The correspondences of this transcription with the IPA are as follows:

a a v v gh fl or j dh ffi e e z z y i th ` i i k k or c l l m m n n x ks o o p p r r s s t t u i f f h x ps ps w o au av or af ou u eu ev or ef ghk g

Recommendations 34 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R 7.10 Application to Swedish The morphosyntactic description of Swedish language has been contributed by Pernilla Danielsson and Jerker J"arborg (Danielsson and J"arborg 1995).

The tags and values used in the application follow the Danish description as the two languages are closely related. The tables are arranged suitable for Swedish but can easily be rearranged if necessary for a comparison among languages.

7.11 Application to Catalan Sections on Catalan have been contributed by Joan Soler (Soler 1995).

The application to Catalan takes into account the lexicon and the corpus achieved in the project Diccionari del Catala Contemporani (DCC). When speaking of the catalan lexicon, the author refers to the DBI (Diccionari Basic Informatitzat - Computerized Basic Dictionary), that has been used, and partially constituted, through the lemmatization tasks of the CTILC (Corpus Textual Informatitzat de la Llengua Catalana). The column devoted to the Catalan tag in the tables displays the actual tag used in the DBI and in the CTILC tagset to represent the attributes and the value.

7.12 Application to Irish The application to Irish has been written by Eoghan Mac Aog'ain (Mac Aog'ain 1995).

Since Irish, like the other Celtic languages, differs in some striking ways from the Romance and Germanic languages of Europe, it is of interest to see how it can be incorporated into the EAGLES framework. The major morphosyntactic peculiarities of Irish are: (1) VSO word order, (2) general complementizers "a" and "go" (somewhat similar to Greek "pou"), (3) person, number and gender attributes on prepositions (as in Hebrew), (4) a null value for person on verbs in all tenses and moods, the so-called free verb, (5) a vocative case for proper nouns (again, as in Greek), (6) a dual form of the very to be, "is" and "t'a", (7) initial mutations of two kinds, lenition and eclipsis (sometimes signalled as "+L" and "+E" in the application), and (8) a velar-palatal (broad-slender) contrast for all consonants, a linguistic feature more common in the East of Europe than in the West. The initial mutations and the palatalization of final consonants are important parts of the morphosyntactic system. Orthographically, lenition is indicated by adding "h" to the initial consonant, and eclipsis by prefixing the mutated consonant to the base consonant. This gives word-initial sequences "bp, dt, gc, mb, nd, ng, and bhf". Palatalization of final consonants is indicated by inserting an "i" before them. The quality of other consonants is indicated in a similar way by the adjoining vowels, "i" and "e" for palatal, "a, o and u" for velar, which explains the large ratio of vowels to consonants in the written language.

8 Present Status of the Morphosyntactic Proposal The present document reflects the results obtained after different phases and many cycles of work.

Recommendations 35 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R - The first version (December 1993) of the document contained the survey of the various systems, their comparison and a first sketch of the consensual nucleus of morphosyntactic specifications. - This preliminary proposal was applied to a number of European languages: Italian, German, Spanish, English, Dutch and French. This phase, which has been developed in an interval of a few months - corresponding to different versions of the present document -, generated a lot of feedback, especially at the language-specific level. - A revision of the preliminary proposal in the light of the feedback coming from the first cycle of applications, which caused changes mainly in the tables at the Level-2b, was performed (March 1994 and June 1994). - The MULTEXT experience (see 2.6.1), in the multilingual framework of a concrete project, has been an important test-bed for the proposal which has brought some modifications in the common proposal itself (October 1994). - More than 20 European and national projects made extensive use of EAGLES specifications, actively interacting with EAGLES: LRE DELIS, RENOS, CRATER, MECOLB, MULTEXT, COPERNICUS MULTEXT-East and TELRI, MLAP PAROLE, ESPRIT-ELSNET, French GRACE, German Textcorpora und Erschliessungswerkzeuge, LE-SPARKLE, ELRA, EUROWORDNET, PAROLE. - Some small revisions were made (May 1995). - Within the MLAP-PAROLE project, the applications to some EU languages not yet covered, i.e. Catalan, Irish and Swedish were contributed and added to the present document, together with the experience gained within the project itself. - Furthermore, comments coming from experts in the field who evaluated these recommendations were incorporated in the present version of the document. - Finally, - as already mentioned above -, the experience gained within the Validation Phase, where Guidelines providing definitions and explicit criteria for the application of the specifications in the different languages were carried out, has been taken into account.

Experience teaches that the process of consensus building in order to arrive at a stable and broadly accepted standard, is of necessity a slow process: the proposal must undergo many phases of discussions between experts in the field and many cycles of applications for testing, evaluating and refining the specifications. It is essential that the proposal/applications interaction proceeds in both directions, as it happened for this document. The responsibles of the document will therefore be grateful for any relevant comment generated by the circulation of the present proposal. The feedback will be taken into account and incorporated in further revisions.

Recommendations 36 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R 9 Noun

N Type Gend Numb Case Count Defin Inflect Decl

M com m sg nom cou

u prop f pl gen mass

l n dat t acc

voc G com m sg

e prop f pl n A * m sg

l f pl D n

N com m sg nom

E prop f pl gen R m+f s+p dat

C acc

bas L com m sg nom cou

e prop f pl gen mass e n dat c c acc h voc

bas

L0 N O U N

com m sg L prop f pl

1 n

L nom cou

2 gen mass a dat

acc It c It n Gr voc Da def Da/Ge weak Ge norm L Du f(m) Gr ind Da indf Da/Ge strg Ge adj

2 Du cont Da unmk Da/Ge mix b Sp trns

Sp notr

Recommendations 37 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R 9.1 Comments Nouns are commonly recognized by the systems under analysis, and an immediate core of agreement as to their grammatical features emerges, as is evident from the table above.

9.1.1 Type This attribute is used to distinguish common and proper nouns. MULTILEX encodes this information among the "syntactic attributes of the LU (Lexical Unit)". AlethDic splits the nouns into two different categories: "nom" and "nom propre" (this diverging treatment is signalled in the table with an asterisk). However, no transduction (see Monachini, Oestling 1992b) and convergence problems arise between these two approaches, since they are easily intertranslatable.

9.1.2 Number and Gender: The "common" and "invariant" values The values "common" - c - and "invariant" - n -, presented in the Italian and Spanish tables -, for the attributes Number and Gender respectively, raise an important issue5: they are not a gender or a number, but they are a signal that gender or number cannot be determined from the form of the word alone. Ontologically they have a quite different status from "m" and "f". They can be seen as a metalinguistic device, or as a third value: it can be seen as a sort of a multilabel whose semantics is a "disjunction" operation: in fact, the value e.g. n is a macro for sg j pl. It is used for the lexicon encoding of the series of the Italian invariant nouns. For these nouns, the number can be decided in the context only on the basis of syntactic considerations, i.e. by observing the agreement with other elements, articles, verbs, adjectives: Parigi e' una citta' europea; le citta' europee Sometimes the number can remain undecided, notwithstanding the context, i.e. when the noun agrees with invariant adjectives. The same stands for the common gender: un bravo insegnante, una ottima insegnante.

9.1.3 Case A discrepancy seems to concern this feature; but "Case" does not appear in the GENELEX and AlethDic models amongst the features for Nouns, owing to their monolingual French-based orientation.

As already discussed in NERC (see Monachini and Oestling 1993b), the values proposed under Case are clearly not mutually disjunctive: rather, some of them overlap (see Introductory part 1.3). Given these overlappings, some facts have to be pointed out:

ffl the values can never appear all together in one language, but have to be associated with

a list of permitted values for that particular language

5We are particularly grateful to Gerald Gazdar and Lynne Cahill of Cognitive and Computing Sciences, University of Sussex for having underlined this issue that has been dealt with in detail and deepened during the EAGLES validation phase in the Guidelines produced for English, French, German and Italian.

Recommendations 38 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R

ffl the signification of a value has to be seen in relation to the other values of the same

language

The relationship between the values, as it can be derived from their use in the analysed tagsets, is illustrated in the following tree:

Case -- /" -- -- oblique -- -- -- -- -- -- -- -- gen nom dat acc

It must be stressed that each language system will use its own pertinent set of values. `Oblique', discussed here as a value of the feature Case, is not pertinent to the category Noun; it is a value marked for Pronouns.

9.1.4 Countability This attribute, which MULTILEX encodes at the syntactic level, is pertinent to many European languages; it is, therefore, here included on level 2a.

9.1.5 Values on Level 2b - language specific On the level 2b of language-specific features the following values and attributes appear (for the meaning and the discussion of the language-specific values, see under the relevant applications): - for Gender `common', for Italian; `fem(masc)' and `context', for Dutch; `transart', `notransart' for Spanish; - for Number `invariant'; - for Case `vocative' and `indeclinable' marked in Greek; - Definiteness with the values `def, indef, unmark', for Danish; - Inflection Type with `weak, strong, mixed' is also on this level to deal respectively with this Danish and German characteristics of nouns.

Recommendations 39 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R 9.2 Application to Italian (Dictionary and Corpus) The following attribute-value sets are applied in Italian to Nouns (It. tag: S for "substantive"):

9.2.1 Type

Attribute value It. example It. tag

Type common (il) libro S

proper Mario SP

9.2.2 Gender

Attribute value It. example It. tag

Gender masculine (il) libro S/m

feminine (la) casa S/f neuter common (l')insegnante S/n

The value common, i.e. the same morphological form for masculine and feminine, which is typical of Romance languages, is pertinent both to lexica and corpora: unlike in the lexicon, in a corpus the ambiguity can often - though not always - be resolved by the context.

l'insegnante capace insegna (not possible to disambiguate)

l'insegnante bravo insegna (possible to disambiguate)

The value neuter is not pertinent to Italian. Some nouns present a sort of fossilized neuter suffix, coming from the ancient Latin neuter gender (in particular, plural nouns denoting "parts of the body" such as membra, braccia, ciglia are still present in Italian, but they are not classified as neuter: they are considered as exceptions in the morphological inflexional paradigms for nouns, and are classified as "feminine", alternating with a "masculine" inflexion)6.

9.2.3 Number

Attribute value It. example It. tag

Number singular (il) libro S/ms

plural (i) libri S/mp

l-spec invariant (la/le) attivita' S/fn 6In general, these forms ending in -a alternate with the regular form of the plural: sg. braccio, pl. bracci, braccia; sg. membro, pl. membri, membra; sg. ciglio, pl. cigli, ciglia. The two forms disambiguate, in the plural, the polysemy contained in the noun in the singular.

Recommendations 40 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R In Italian, the value invariant (It. tag: n), i.e. the same morphological form for the singular and the plural, is also used. In particular, this value has to be coded in a lexicon, while in a corpus the context usually solves the ambiguity.

L'uomo svolgeva alcune attivita' particolari (plur.)

L'uomo svolgeva un' attivita' particolare (sing.)

9.2.4 Countability

Attribute value It. example It. tag Countability countable la penna uncountable lo zucchero This feature is neither used in the tagging of the corpus nor is it at present encoded in the Italian lexicon, although its coding is planned.

Application test: "Countable" is the noun which has the plural form; usually it cannot be involved in the partitive construction, ex. due penne but not * un po' di penna; a mass noun admits the partitive, ex. un po' di zucchero, and if used in the plural tre zuccheri means "three different types of sugar".

9.2.5 Features not pertinent to Italian The feature Case does not apply to Italian nouns.

Definiteness and Inflection Type are not pertinent to Italian.

Recommendations 41 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R 9.3 Application to German 9.3.1 Type

Attribute value example tag

Type common (das) Buch NN:Neut.Nom.Sg

proper Hans NE:Masc.Nom.Sg

9.3.2 Gender

Attribute value example tag

Gender masculine (der) Mann NN:Masc.Nom.Sg

feminine (die) Frau NN:Fem.Nom.Sg neuter (das) Haus NN:Neut.Nom.Sg

9.3.3 Number

Attribute value example tag

Number singular (der) Mann NN:Masc.Nom.Sg

plural (die) M"anner NN:Masc.Nom.Pl

9.3.4 Countability This feature is not used in the IMS-Tagset for German.

9.3.5 Case

Attribute value example tag

Case nominative (der) Mann NN:Masc.Nom.Sg

genitive (des) Mannes NN:Masc.Gen.Sg dative (dem) Mann NN:Masc.Dat.Sg accusative (den) Mann NN:Masc.Akk.Sg

9.3.6 language-specific features In German, there are some nouns with adjectival inflection (especially nominalized adjectives). These nouns distinguish strong, weak and mixed inflection, depending on the preceding determiner (cf. section 11.3.6).

Attribute value example tag

Inflection strong (nichts) Gutes NN:Neut.Nom.Sg.St

weak (der) Beamte NN:Masc.Nom.Sg.Sw mixed (ein) Beamter NN:Masc.Nom.Sg.Mix

9.4 Application to English English nouns do not have gender as a morphosyntactic category. Also, case is a dubious category, since the only basis for a case distinction in modern English lies in the 's or s' ending

Recommendations 42 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R attached to nouns and to some pronouns. It is arguable, however, that in modern English, this is not a case form, but an enclitic postposition. (This would explain the occurrence of phrases such as in a month or two's time or in someone else's garden, where the 's is clearly suffixed not to the head noun, but to the whole phrase.) Hence, in the present tagset, case is not applied to nouns, and the 's is treated as a postposition.

9.4.1 Type

Attribute values Examples Tags Type common book NCs

proper Martha NPs

9.4.2 Number

Attribute values Examples Tags Number singular book NCs

plural books NCp

Recommendations 43 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R 9.5 Application to Dutch 9.5.1 Type

Attribute value Example Tag

Type common boek N

proper Piet N

CELEX has no specific attribute-value tag for Proper Nouns as a separate category, but has a proper noun subclassification tag set distinguishing four values. So each Proper Noun gets two tags: `N' for Noun and `geo.' for the type of Proper Noun geographical names, and so on: `N' and `pers.' (names of people), `N' and `merk.' (Company or brand names), `N' and `over.' (other).

Attribute value Example Tag Proper-Noun-Type geographic Belfast N geo.

persons Wilma N pers. brand Droste N merk. other Teleac N over.

The proposed attribute values Noun and Proper Noun are entirely appropriate on the first level for Dutch as well. The Proper Noun subclassification would be introduced as an L2 tag.

9.5.2 Gender CELEX distinguishes five basic gender tags. Apart from these, a number of combinations of these five tags are also possible (double tags).

Attribute value Example Tag

Gender masculine wijn m.

feminine ruimte v. neuter vertrek o. fem.(masc) bierhal v.(m.) context gelovige m.-v.

The fourth category is: female nouns which can be treated as male. The fifth category is: nouns whose gender is context-dependent. These are, for example, Adjectives and Present and Past participles used as nouns. Referential characteristics, such as possessive nouns in the near context or other contextual factors must be used to help decide on the respective gender:

Recommendations 44 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R

De gelovige heeft een kaars in HAAR handen The worshiper holds a candle in HER hands

`gelovige' (worshiper) is feminine.

De gelovige heeft een kaars in ZIJN handen The worshiper hold a candle in HIS hands

`gelovige' (worshiper) is masculine.

See also the subsection Language-specific features.

9.5.3 Number

Attribute value Example Tag

Number singular kanarie e

plural kanaries m

These attribute values are not present in the LEMMAS lexicon, but are presented in the CELEX Wordforms Lexicon. As will be seen below, `e' and `m' also feature in verbal tags.

CELEX distinguishes two other Noun features as well: diminutive singular and diminutive plural. Flection Type tags are : `de' and `dm'.

Attribute value Example Tag Diminutive-Number singular kanarietje de

plural kanariestjes dm

This could be an L2 tag.

9.5.4 Countability CELEX does not account for countability, but it is pertinent to Dutch.

Recommendations 45 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R 9.5.5 Case CELEX does not treat Case as part of the syntactical system but as part of the inflectional system. In Dutch the genitive `s', directly fixed behind the noun, marks possession, as in English. It also substantivizes adjectives: (iets) moois, (iets) or (het) lekkers etc. Case, other than genitive, is not pertinent to modern Dutch. Only rchaic forms have the former case-dependent inflection. This must be the reason why case values are not given in the Syntactical part of the CELEX Lemmas Lexicon but are given in the Wordforms lexicon.

Attribute value Example Tag

Case genitive (dag des) oordeels G

dative (te) gronde (richten) D

Both tags are always combined with a `singular' or `plural' tag. There are also GP tags (Genitive of Adjective Positive). Some morphologists count such word forms as adjectives, others as nouns. We count them as nouns.

Ge: Genitive singular: hoofds. GP: Genitive Positive: lekkers, moois, nieuws (nouns deriving from adjectives). Gm: Genitive plural: only pronouns in the CELEX lexicon! De: Dative singular: bate, behoeve, berde etc. Dm: Dative plural: only pronouns in the CELEX lexicon!

See also Case under Articles and Pronouns.

9.5.6 Language-specific features Concurrent Gender distinction Since in Dutch the distinction between the male and female gender of nouns is disappearing, another appropriate gender distinction is the distinction neuter/non-neuter, which is expressed in the definite article determining the noun: `de' for female/male words and `het' for neuter words. As a consequence of the fact that some words can be preceded by both articles, CELEX distinguishes three tags:

Attribute value Example Tag De/Het Woord non-neuter de deur de

neuter het schip het de or het de/het jolijt de/het

We need a special tag to mark up nouns which are part of a separable verb: `paard' in `paardrijden', separated word form: `Ik heb paard gereden'. And `hout' in `houthakken', separated word form: `Wij hebben hout gehakt' etc. See also under Separability and compare language-specific features of Adjectives and Adverbs.

Recommendations 46 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R 9.6 Application to Spanish 9.6.1 Type

Attribute Value Example Tag Type common libro common

proper Pedro proper

9.6.2 Gender

Attribute Value Example Tag Gender masculine (el) libro masc

feminine (la) casa fem neuter common (el/la) responsable * l-spec transart el area yes

The value neuter is not pertinent to Spanish: we have no neuter nouns. There are nouns in Spanish which could be valued as common since they can be both masculine and feminine. However, in Eurotra there is no such value: instead we leave the attribute "gender" un-valued at the lexical level. It is the grammar that fills this value via unification, that is, any noun with its "gender" attribute un-valued will unify with both feminine and masculine articles. A special case in Spanish with respect to gender is that of feminine nouns beginning with stressed "a" (having "h" before or not). These nouns take, when in the singular, the masculine allomorph of the (un)definite article if this immediately precedes the noun:

el aguila calva (the(masc,sing) bald(fem,sing) eagle) las aguilas calvas (the(fem,plu) bald(fem,plu) eagles)

la gran aguila (the(fem,sing) great eagle) In order to avoid problems of concord, we add a new boolean attribute ("transart") which serves to distinguish these nouns: when its value is `yes', a formation rule performs the marking of the article in order for the translator to change its gender value:

Attribute Value Example Tag transart yes el area yes

no el ni~no no

This "transart" attribute could be treated as a language-specific value for the attribute "Gender".

Recommendations 47 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R 9.6.3 Number

Attribute Value Example Tag number singular el libro sing

plural los libros plu

A special case concerning "number" is that of those nouns with the same morphological form for the singular and the plural: eg: la crisis, la dosis, el martes, el chasis, el an'alisis

las crisis, las dosis, los martes, los chasis, los an'alisis

Just as in the case of "el/la responsable" above, we leave these nouns unvalued for "Number". We get the value via unification with the value of the article and/or adjective accompanying the noun. Of course, we could add a language-specific value to mark these nouns. Another special case is that known as "pluralia tantum". This group comparises nouns which are always plural, where the plurality does not come from any inflectional process but is rather inherent to the noun itself: eg: v'iveres, a~nicos, entendederas... (provisions, bits, brains)

9.6.4 Case Not applicable to Spanish nouns although it is present in the Eurotra dictionary for nouns since we include pronouns in the noun category. To distinguish between nouns and pronouns we have the attribute "Nform":

Attribute Value Example Tag nform normal casa norm

clitic le cli pronoun 'el pro

This requires us to include new attribute referents to pronouns such as "possessive" in our category Noun.

9.6.5 Countability This attribute is not included in Eurotra's lexicons at the morphosyntax level. Countability is treated as a semantic attribute with the values "countable" and "uncountable".

Recommendations 48 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R 9.7 Application to French (Corpus) 9.7.1 Type

Attribute value Example Tag Type common livre SUBSMS

proper UNIX NPRO

9.7.2 Gender

Attribute value Example Tag Gender masculine homme SUBSMS

feminine femme SUBSFS

Note that some proper nouns can bear gender information (see below). 9.7.3 Number

Attribute value Example Tag Number singular b'eb'e SUBSMS

plural b'eb'es SUBSMP

Note that some proper nouns can bear number information (see below). 9.7.4 EAGLES features not applicable Countability is not applicable to the IBMF tagset. Being clearly of a semantic nature, it is not obvious why it should feature in a morphosyntactic description scheme. Case, Definiteness, Inflexion type are not applicable to French.

9.7.5 IBMF Tagset features not applicable in EAGLES In the IBMF tagset, it has been found useful for disambiguation purposes to refine the proper noun class into subclasses. It can be noted that some of these classes bear gender/number information.

Tag Meaning Examples XFAMIL family name Smith, Dupont XPAYFP country name, fem.pl. (les) Seychelles XPAYFS country name, fem.sing. (la) France XPAYFP country name, masc.pl. (les) U.S.A. XPAYFP country name, masc.sing. (le) Danemark XPREF christian name, fem. Mary, Marie XPREM christian name, masc. John, Jean XSOC company name Olivetti, IBM XVILLE city name Paris, Pise

Recommendations 49 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R 9.8 Application to French (Lexicon) 9.8.1 Type ------------ ----------- ----------- ---- Attribute Value Example Code ------------ ----------- ----------- ---- Type common livre c

proper Jean p ------------ ----------- ----------- ----

9.8.2 Gender ------------ ----------- ----------- ---- Attribute Value Example Code ------------ ----------- ----------- ---- Gender masculine homme m

feminine femme f ------------ ----------- ----------- ----

9.8.3 Number ------------ ----------- ----------- ---- Attribute Value Example Code ------------ ----------- ----------- ---- Number singular homme s

plural femme p ------------ ----------- ----------- ----

9.8.4 Case Non applicable to French.

9.8.5 Combinations --------- ----------- Tag Example --------- ----------- Ncms- homme Ncmp- hommes Ncfs- femme Ncfp- femmes Npms- Jean Npmp- Pays-bas Npfs- Anne Npfp- Pyrenees --------- -----------

Recommendations 50 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R 9.9 Application to Portuguese The following attribute-value sets are applied in Portuguese to Nouns :

9.9.1 Type ----------- --------------- -------------- ------------- Attribute Value Example Tag ----------- --------------- -------------- ------------- Type common rapaz

proper Lisboa ----------- --------------- -------------- -------------

9.9.2 Gender ----------- --------------- -------------- ------------- Attribute Value Example Tag ----------- --------------- -------------- ------------- Gender masculine rapaz

feminine rapariga neuter common ----------- --------------- -------------- -------------

The value common is not explicitly used in GENELEX, though the nominal inflectional patterns built in the framework of this project implicitly make use of it for nouns such as, for instance, (o/a) dentista. For Portuguese this value can be a L2a value.

The value neuter is not pertinent to Portuguese. (Although, as in other languages, some nouns from classical languages presenting a kind of neuter suffix [e.g. lexica] can be used in Portuguese, they are typically assigned a pattern of masculine inflection.)

9.9.3 Number ------------- ------------- --------------- ------------------ Attribute Value Example Tag ------------- ------------- --------------- ------------------ Number singular rapaz

plural rapazes invariant ------------- ------------- --------------- ------------------

The value invariant is not explicitly used in the Portuguese GENELEX model, though the nominal inflectional patterns built in the framework of this project implicitly make use of it

Recommendations 51 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R for nouns such as, for instance, (o/os) atlas, (a/as) sandes. For Portuguese this value can be a L2a value.

9.9.4 Countability

Though this feature is not encoded in the GENELEX morphological layer, it is true that nouns in Portuguese can be used sometimes in a "count" sense and other times in a "mass" sense. Considering the principles of granularity supporting the distinction between L1 and L2a, we think that this feature should be considered "Application dependent".

9.9.5 Features not pertinent to Portuguese The features Case, Defineteness and Inflection Type don't apply to Portuguese nouns.

Recommendations 52 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R 9.10 Application to Danish For Danish nouns the following features are relevant: Type, Gender, Number, Case (restricted) and Definiteness (language-specific property). The attribute `Countability' may also be useful at the lexicon level.

9.10.1 Type

Attribute value Example Tag

Type common lampe sb

proper Peter sb prop

9.10.2 Case The Danish nominal inflectional system comprises only two cases - genitive and non-genitive (oblique) - which are mutually disjunctive. The value `ngen' will not be used by the corpus tagger, because this oblique form is unmarked in Danish.

Attribute value Example Tag

Case non-genitive lampe sb sg ngen

genitive lampes sb sg gen

9.10.3 Countability This feature is treated within the EDEMD and within traditional Danish dictionaries as defective inflectional patterns like `singularia tantum' and `pluralia tantum', as a parallel to countability, i.e. non-countable or mass words like `maelk' (milk), `vand' (water), and a number of deverbal nouns like `drift' (drift) are non-countable and also singularia tantum. Those words do not have a plural form. On the other hand, words like `penge' (money) only have a plural form, which is also marked in the dictionaries mentioned above. The tags below are examples of the Eurotra encoding, and they are values of the attribute sub-category of nouns in the lexicon. Other useful tags would be `count' and `non-count' as values in the lexicon for the attribute countability. The corpus tagger will recognise the inflected forms occurring in the corpus.

Attribute value Example Tag Countability singulare tantum (en) kulde sb singtant

plurale tantum penge sb plutant

9.10.4 Gender

Attribute value Example Tag

Gender common (en) lampe sb comm

neuter (et) bord sb neut

Recommendations 53 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R 9.10.5 Number

Attribute value Example Tag

Number singular (en) lampe sb sg

plural lamper sb pl

9.10.6 Language-specific features: Definiteness In Danish, nouns can be marked for definiteness with the enclitic article; in singular common gender: -(e)n; singular neuter: -(e)t; and in plural: -(e)ne (no gender distinction). The order of inflectional endings is: number + enclitic article + genitive suffix, e.g. lampe + s; lampe + r + ne + s, which give the tag combinations sb comm sg indef gen and sb comm pl def gen, respectively.

Attribute value Example Tag Definiteness indefinite (en) lampe sb comm sg indef

definite lampen sb comm sg def

Recommendations 54 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R 9.11 Application to Greek Nouns in Greek are inflected for number and case. The following set of attribute-value pairs is derivable from information coded in the Moprhological Lexicon (either in the stem or the inflectional paradigm):

9.11.1 Type

Attribute value Gr. example Gr. tag

Type common vivlio NoCm

proper Ghiannys NoPr

9.11.2 Gender

Attribute value Gr. example Gr. tag

Gender masculine ouranos NoCmMa

feminine karekla NoCmFe neuter vivlio NoCmNe l-spec masc-fem tamias NoCmMf

The value masc-fem is not used in the current codification of the ILSP lexicon. However, it is considered valuable information and is planned to be coded. This value characterises nouns that share the same form both in the masculine and the feminine gender (e.g. nouns denoting professions); note that given that these nouns do not have a neuter gender, the label masc-fem is considered preferrable to common (which is a value used for the attribute of gender, but not explained in this section as it is used for adjectives mainly). The linguistic context can contribute to the resolution of the ambiguity in certain cases (usually on the basis of the modifying article and/or adjective, given that they must agree in gender with the noun); however, for reasons of uniformity, when the value is adopted in the coding system, the value will remain during corpus annotation as well:

O tamias einai apasholymenos (masc.)

Y tamias einai apasholymeny (fem.) but Oi tamies efughan (masc. or fem.)

9.11.3 Number

Attribute value Gr. example Gr. tag

Number singular vivlio NoCmNeSg

plural vivlia NoCmNePl

l-spec invariant asanser NoCmNeNv

Recommendations 55 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R The value invariant is used in Greek for those nouns that do not have a morphological variant for the singular and plural number. This holds mainly for foreign words that have been incorporated in the Greek language and have been accepted as "Greek words" but which, however, have not been adapted to the Greek inflectional system. The linguistic context (modifying article/adjective) can serve as a disambiguating device in the case of corpus tagging:

To asanser vrisketai sto isogheio (sing.) Einai halasmena kai ta dhuo asanser (plural)

9.11.4 Case

Attribute value Gr. example Gr. tag

Case nom vivlio NoCmNeSgNm

gen vivliou NoCmNeSgGe acc vivlio NoCmNeSgAc dat (suneheia) NoCmNeSgDt l-spec voc Ghianny NoPrMaSgVo l-spec indcl asanser NoCmNeNvIc

As regards the value dat, we must report that its inclusion in our tagset was forced by the needs of corpus annotation.

The problem arises from the phenomenon of diglossia in Modern Greek, i.e. the parallel existence of types of "dimotiki" and "katharevousa" at the same time in the language. Although "dimotiki" is the official language of Greece since 1974, a lot of types of katharevousa are actively used, mainly in the written language.

Thus, as regards the dative case, it is used in the ILSP tagset although it does not belong to the current inflectional system of Modern Greek. Only four case values exist at present: nominative, genitive, accusative and vocative. However, there remain in the Modern Greek language certain word forms from "katharevousa" (mainly fixed phrases), which make use of this case value, thus enforcing us not only to recognise them but also to characterise them. These word forms are mostly used in fixed phrases; in the future, we plan a mechanism that will be able to recognise and tag appropriately "groups of words", i.e. two or more words functioning together as one (e.g. complex adverbs, complex prepositions, fixed expressions, etc.):

en suneheia (complex adverb, consisting of a prep. and a noun in the dative case) The value indcl, mutually exclusive with the other values, is used in Greek for words that retain the same form in whichever case they are found, either in the singular or in the plural number. As commented earlier, this value mainly applies to foreign words which have been incorporated in the Greek language without having taken the morphological characteristics of the Greek inflectional system. The case value could be disambiguated taking into account the linguistic

Recommendations 56 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R context in corpora, on the basis of agreement features, although, for reasons of uniformity and in order to avoid further complication of the disambiguation process, at present it is not disambiguated:

To asanser vrisketai sto isogheio (nominative) Vryka tyn porta tou asanser anoihty (genitive)

The vocative case (value voc) is rarely used in written texts; it can only be found in literature, and, in general, in oral or written dialogues. The sentence in which it is found is often followed by an exclamation mark:

Ghianny, ela edhw! The above two values, indcl and voc are specific to the Greek language, and, therefore, belong to level L2b.

It is often the case (as regards nouns of the feminine or neuter gender) that the same form is used for both the nominative and the accusative case. Although the use of a new value, nom-acc, would contribute to economy in the lexicon, we have decided to keep them as distinct values and proceed to ambiguity resolution in the process of corpus tagging. Disambiguation can only be made on the basis of the linguistic context, either simply by inspection of the modifying article/adjective, or, if that is not sufficient, by resort to shallow syntactic analysis (subject/object role):

Y karekla einai sto dhiplano dhwmatio (nom.)

Koitazei tyn karekla (acc.)

but To vivlio vrisketai sto trapezi (subj. - nom.)

Vryka to vivlio (obj. - acc.)

9.11.5 Countability

Attribute value Gr. example Gr. tag Countability countable vivlio mass plythos This feature is not used at present in the Greek Morphological Lexicon (and, thus, is not used in the corpus tagging either). However, it applies to the Greek linguistic system and is planned to be coded in the future.

9.11.6 Features not pertinent to Greek Features Definiteness and Inflection Type are not pertinent to Greek.

Recommendations 57 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R 9.12 Application to Swedish The following features are relevant for the Swedish nouns; Type, Gender, Number, Case and the language specific property Definiteness.

Type Gender Number Definiteness Case proper ordinary

genitive common nonneuter singular indefinite

neuter plural definite

9.12.1 Type This is a lexical feature. The values proper and common are pertinent to Swedish and govern certain other features; cf below.

Attribute value Example Tag Type common bord sb

proper Birgitta sb-prop

9.12.2 Gender The features is normally not applicable to proper nouns. For common nouns, the feature takes the values nonneuter and neuter. The gender values are lexically determined and govern the realisations of the features Definiteness and (partially) Number.

Attribute value Example Tag Gender nonneuter (en) stol sb-nonneu

neuter (ett) bord sb-neut

9.12.3 Number This feature is normally not applicable to proper nouns. For common nouns, the values are singular and plural. The value plural has six different types of realisations, depending on the declination of the noun, which is lexically determined and partially dependent on the Gender value. In one of these declinations, mainly containing neutral nouns, plural is unmarked and thus morphologically indistinguishable from singular.

Attribute value Example Tag Number singular (en) stol sb-sg

plural stolar sb-pl

Recommendations 58 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R 9.12.4 Definiteness This is a language-specific morphological feature, since Swedish, like the other Scandinavian languages (and certain Balcan languages) possesses enclitic definite articles. The feature is normally not applicable to proper nouns. The values are indefinite and definite and their realisations is directly dependant on the Gender value if the Number value is singular. In plural, the Gender distinction is unmarked.

Attribute value Example Tag Definiteness indefinite (en) stol sb-nonneu-sg-indef

definite stolen sb-nonneu-sg-def

9.12.5 Case For both noun Types, this feature takes the values ordinary and genitive. (For certain pronouns, ordinary is split into nominative and oblique.) The value genitive is almost universally realised by the suffix -s.

Attribute value Example Tag Case ordinary stol sb-sg-ord

genitive stols sb-sg-gen NB: The features Number, Definiteness and Case are realised (when at all) in just that order; thus:sb plural definite genitive

stol -ar -na -s

Recommendations 59 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R 9.13 Application to Catalan The two main tags applied in catalan to Nouns are M and F, depending on the gender of them. This allows us to differentiate in the lemma's level the gender of nouns (which is considered as inherent to the category) and the gender of adjectives (which is marked as a characteristic of the form). An automatic change for these two tags in a common one could be possible.

9.13.1 Type

--------------------------------------------------------

Attribute value Ca. example Ca. tag --------------------------------------------------------

Type common (el) llibre M

(la) casa F --------------------------------------------------------

proper Josep PN --------------------------------------------------------

9.13.2 Gender

--------------------------------------------------------

Attribute value Ca. example Ca. tag --------------------------------------------------------

Gender masculine (el) llibre M --------------------------------------------------------

feminine (la) casa F --------------------------------------------------------

neuter --------------------------------------------------------

common l'amor MF/s

aquest amor MF/ms aquesta amor MF/fs --------------------------------------------------------

We apply the value common in a very restrictive way. Nouns as ensenyant, artista, can be (always) disambiguated by the context (but in some cases not the immediate context), because they refer to entities that have sex (and this is obviously determinant for the grammatical gender of the nouns). But there are other cases of nouns such as amor, mar, vigor that have, for etymological reasons, two possibilities of gender, and some of them can not be disambiguated by the context. The value neuter is not pertinent to Catalan.

9.13.3 Number

--------------------------------------------------------

Recommendations 60 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R

Attribute value Ca. example Ca. tag --------------------------------------------------------

Number singular (el) llibre M/s

(la) casa F/s --------------------------------------------------------

plural (els) llibres M/p

(les) cases F/p --------------------------------------------------------

There are some invariant nouns suchs as (el/els) llapis, but they are not marked differently because in the corpus the context always solves the ambiguity. There are also cases of pluralia tantum (like tisores), which are tagged MP or FP as the code of lemma.

9.13.4 Countability This feature could be pertinent in the tagging of the corpus, but it is not used in the tagset for Catalan.

9.13.5 Features not pertinent to Catalan The features Case, and the more language-specific ones Definiteness and Inflection Type, are not pertinent to Catalan nouns.

Recommendations 61 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R 9.14 Application to Irish 9.14.1 Type =========================================== Attribute values Examples Tags =========================================== Type common leabhar [book] NC

proper Maighre'ad, Sea'n NP ===========================================

(As no established tagset exists as yet for Irish, the tags given here and below are provisional.)

Recommendations 62 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R 9.14.2 Gender ======================================== Attribute values Examples Tags ======================================== Gender masculine leabhar [book] NCm

feminine teanga[language] NCf ========================================

9.14.3 Number ======================================== Attribute values Examples Tags ======================================== Number singular leabhar NCs

plural leabhair NCp ========================================

9.14.4 Case ================================================== Attribute values Examples Tags ================================================== Case nominative leabhar NCNs

genitive leabhair NCGs vocative A Mhaighre'ad/Shea'in NPV ==================================================

Like Greek, Irish has a vocative case. It requires (1) insertion of vocative particle, (2) lenition of initial consonant, and, in the case of masculine proper names, (3) palatalization of final consonant. The dative was recently dropped from the written language, except for some stock phrases, but it survives in speech.

9.14.5 Features not pertinent to Irish Countability and definiteness have no useful application in Irish. The plural is an obligatory entry in Irish lexica, because of its irregularity, and its absence could only indicate uncountability.

9.14.6 Language specific features: Contrast ===================================================== Attribute values Examples Tags ===================================================== Contrast unmarked (mo) leabhar [(my) book] NCms

marked (mo) leabharsa [(my) book] NCmsC =====================================================

Recommendations 63 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R Common nouns may carry a suffix to intensify the value of the Person attribute of a preceeding pronominal determiner. (See below for the classification of possessive pronouns/adjectives as determiners.) The suffix, like the determiner, is declined for person, number and, in the third person, gender. Mo leabharsa would be similar to the English my book with a contrastive intonational emphasis on the my. However, if intonational emphasis is used in Irish, it must fall on the noun, leabhar, not on the determiner or on the contrastive suffix.

The attribute Contrast also applies to pronouns, personalised prepositions, and composite (verbpronoun) forms of the verb.

Recommendations 64 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R 10 Verb

V Type Fin Vf-M Tns P N G Asp Vce Refl MVf AuF Aux Sep Clt Zu M mai inf ind pres 1 s ifve act LU LU

u mod prpt sub fut 2 p perf pas

l prf pspt impr past 3 pfve refl t prg ger prog norf

i pas sup l cop cond G ind pres 1 s m stat pass etr

e sub impf 2 p f proc no-p avr n impr fut 3 resul

e cond past

l part pa-s

inf A ind pres 1 s m

l sub impf 2 p f e impr fut 3

t cond past h part pa-s

inf N inf ind pres 1 s m

E part sub impf 2 p f R ger impr fut 3 sp mf

C sup cond past

fin pret L mai inf ind pres 1 s ifve act hv sep

e mod part sub impf 2 p simp pas be oth e prf ger impr past 3 refl c prg sup cond fut h pas fin

pph

L0 V E R B

E mai fin ind pres 1 sg m A aux sub impf 2 pl f G mod impr fut 3

cond past no- inf L fin part

1 ger

sup L s-aux pfve act rfl trs prg

2 cop ifve pas norf int prf a imp pss

pph L En ing It c hv Du sep clt Ge inf

2 En bas be Du nos nocl Ge zu b

Recommendations 65 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R 10.1 Comments Verbs are consensually recognized at the level of category.

The major problem concerns the different internal organizations of the verbal systems which characterize Romance languages on the one hand as opposed to English on the other hand. Romance languages, but also, e.g., German, are highly inflectional, whilst English presents very few inflections and scarce morphological distinctions for moods and tenses.

Furthermore, the Leech/Wilson and MULTILEX proposals, in particular, introduce new distinctions which in the present proposal appear distributed on different levels (see sections 3.1.3 and following).

10.1.1 Type This attribute permits the encoding of the nature of a verb as `main', `auxiliary', `modal' (distinctions seen as recommended), and `copulative' and `semiauxiliary' (optional, and therefore proposed at level 2a, e.g. venire in Italian is used in some cases as a variant to essere to form the passive).

10.1.2 Finiteness, Verb-Form/Mood A different grammatical tradition motivates in English the use of the feature Verb-Form, with values which in Romance languages are typical of the feature Mood. This emerges immediately from the above table, if MULTILEX, NERC and the Leech/Wilson proposal, all designed for a multilingual encoding system, are compared with GENELEX and AlethDic, which are modeled for French verbs, i.e. a system belonging to the Romance tradition.

In a certain sense, the matter can be seen as a problem of the distribution of some values under one or another attribute. As already noted in the NERC survey (Monachini and Oestling 1992b, p.13), the feature Verb-Form appears "strange" from a Romance language point of view, and therefore the problem was left open as an area for further analyses and suggestions. The features Verb-Form and Mood in Monachini and Oestling have been arranged in a tree, with Mood hierarchically dependent on one of the possible values of Verb-Form, i.e. "finite".

The proposal formulated in what follows for reaching a consensus is to introduce, besides "finite", the value "non-finite" as a new value of a new feature Finiteness, and to create a feature called Verb-Form/Mood (Vf-M) with the relevant values which could be seen in subordinate position with respect to the values `finite' and `non-finite'. In such a way, the feature Verb-Form/Mood can be considered totally in hierarchical subordination with respect to Finiteness. For this reason, each language-specific morphological system has to define explicit and strong constraints on which combinations of attributes and values are meaningful for the system (see e.g. the Italian application, section 3.2.1). The situation can be represented as follows:

Finiteness

/" non-finite finite

Recommendations 66 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R

-- -- Verb-Form/Mood Verb-Form/Mood

/----" /----" inf part ger sup ind sub impr cond

With this new arrangement, English encoding systems, which do not reach the level of distinction given here for finite Verb-Form/Mood, can stop at the encoding of Finiteness with the value "finite", while Romance languages will not have problems in being compelled to call some Moods Verb-Forms (contrary to their tradition). Furthermore, the above proposal is completely compatible with the Romance verbal systems, since, in those systems, the values of Moods are implicitly finite and non-finite, in the speaker's intuition.

10.1.3 Tense Tense, as can be observed, does not contain values for "compound tenses", in general past tenses, which in corpora are not usually dealt with by automatic taggers. The possibility of "re-building" and deducing analytic verbal forms, by means of e.g. rules and permitted combinations of "auxiliary-past participle" in the various languages, is provided by values presented here as special extensions, which encode the auxiliary nature of a verb. This introduces a distinction between the requirements of an encoding scheme for corpora and for lexica: the latter also need specifications for compound tenses.

Another problem for Tense is posed by the value "past", since it changes its "meaning" in the various verbal systems, depending on the opposition relationships among values in a language. The English `past' does not have the same meaning as the `past' in the Romance languages: it is not opposed to an `imperfect' value, but instead it seems to be in opposition to the `present'. In NERC, a proposal for the disambiguation of this complicated situation was made with the following tree:

Tense / -- " future present past.E

/" past.R (perfect) imperfect

In Romance language systems, the values `past' and `imperfect' are opposed and designate two different aspects of a past action. Both are opposed to the `present' with respect to the notion they represent: `past' is a non-durative action finished in the past and `imperfect' is a durative action initiated in the past. In order to avoid misunderstandings, a tentative solution could be to rename the Romance `past' value as `perfect', since it is opposed to `imperfect'.

10.1.4 Person, Gender and Number These features do not pose problems.

Recommendations 67 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R 10.1.5 Aspect The values proposed here are `perfective' and `imperfective' and, given the syntactic pertinence of this attribute, they appear at level 2a.

10.1.6 Voice The values `active' and `passive' are proposed at the level 2a of optional features. The distinction between `reflexive' vs. `non-reflexive', foreseen by MULTILEX under this attribute, is treated here as a separate feature (see below).

10.1.7 Reflexivity This feature encodes phenomena which in the MULTILEX scheme are collapsed under the feature Voice. The two values of boolean type are arranged on level 2a for the encoding of verbs appearing with the reflexive pronoun as a clitic. The further distinction between true reflexive, pronominal, reciprocal, etc. can be dealt with only at the syntactic level.

10.1.8 Main-Verb Function and Auxiliary Function These two attributes have to be seen in subordinate position with respect to the two values of Type `main' and `auxiliary' and belong to level 2a. Their respective values are:

Main-Verb Function (MVf ): `trans', `intrans', `impers'.

Auxiliary Function (AuF): `prg': for auxiliaries used to form progressive tenses (be, ...) `prf': for auxiliaries used to form perfect tenses (have, avere, ...) `pss': for auxiliaries used to form passives (etre, worden, werden, ...) `pph': periphrastic auxiliary (do, ...)

Auxiliary Function, proposed in the present work, encodes phenomena which in the surveyed schemes are collapsed under the feature Type. Main-Verb Function is a completely new feature.

10.1.9 Level-2b specifications Auxiliary This attribute encodes information concerning the choice of auxiliary for perfect or compound tenses and obviously is language-specific (e.g. have, be for English).

Separability This is used in Dutch for verbs including separable particles, such as faengt ... an.

Clitic This feature with boolean values is introduced at language-specific level to encode the presence in the verbal form of the clitic.

Recommendations 68 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R language-specific values for Tense The values proposed here for English are `ing-form' and `base-form`. The value `infinitive incorporating "zu"' is added for German.

Recommendations 69 May, 1996 EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R 10.2 Application to Italian The following set of morphological features are pertinent to Verbs in Italian: Mood, Tense, Person, Gender, Number. These features, variously combined, dynamically generate the inflected forms of the Italian verbal system, but not all the features and sometimes not all the values of a feature are applied to an inflected form. The constraints in the application of these features are made explicit in the sections devoted to each feature.

The following attribute-value sets are applied to Italian verbs. 10.2.1 Type The verbs avere and essere are marked as auxiliaries. It should be noted that an automatic tagger is not able to disambiguate the cases in which these verbs function as full verbs [il bambino e' a casa (the child is at home), Io ho un cane (I have a dog)] from the cases when they are auxiliaries.

10.2.2 Finiteness and Verb-Form/Mood The feature Finiteness is not encoded in itself, i.e. does not have a special mark in the Italian corpus or in the lexicon. It is implicity contained in the various distinctions of Mood, which is hierarchically dependent on the two possible values of Finiteness: on the one hand, infinitive, participle, gerund are non-finite; on the other hand, indicative, subjunctive, conditional, imperative are finite. It can, therefore, be derived automatically from existing codes. The application of these two features to Italian can be represented as follows:

Attribute value It. ex. It. tag Finiteness non-finite - -

Attribute value It. example It. tag Verb-Form/Mood infinitive amare V/f

participle amato V/p gerund amando V/g supine - -

Finiteness finite - -

Attribute value It. example It. tag Verb-Form/Mood indicative amo V/s1i

subjunctive amasse V/s3c imperative amate V/p2m conditional amerei V/s1d

10.2.3 Tense It has to be said that not all the values listed for the feature Tense are applied to all the values of Mood. The table below represents the constraints on the application of the values of Tense to Mood in Italian. Note that, if a value of tense is applied to a value of mood, but is expressed by a

Recommendations 70 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R compound form, this is noted with (cmpd). All details about compound forms are given below.

Finiteness finite non-finite Verb-Form/Mood indic subj condit imper inf part ger Tense pres amo ami amerei ama amare amante amando

impf amavo amassi fut amero' past amai +(cmpd) (cmpd) (cmpd) (cmpd) amato (cmpd)

Considerations about the "meaning" of past in Italian have to be added: past corresponds here to passato remoto. A problematic aspect of Tense, that of "simple" tenses and "compound" tenses, and the connected difference in requirements for corpora and lexica, has already been dealt with in the general section of comments on the table for verbs. We give here the list of permitted "auxiliary-past participle" combinations in Italian, in order to obtain compound tenses.

Auxiliary Past participle Compound tense It. ex. indic pres past participle passato prossimo ho amato indic impf past participle trapassato prossimo avevo amato indic past past participle trapassato remoto ebbi amato indic fut past participle futuro anteriore avro' amato subj pres past participle congiuntivo passato abbia amato subj impf past participle congiuntivo trapassato avessi amato cond pres past participle condizionale passato avrei amato inf pres past participle infinito passato avere amato ger pres past participle gerundio passato avendo amato

10.2.4 Person

Attribute value It. example It. tag

Person 1 amo V/s1

2 ami V/s2 3 ama V/s3

10.2.5 Gender This feature, among the simple morphological verbal units, is pertinent only to the Mood participle: it agrees with nouns as an adjective (note on the agreement with object avendo letta la lettera, uscii, note on ablativo assoluto). The value common does not apply to the past participle. The gender of past participle, in context, helps to disambiguate the gender of the noun itself, when this is morphologically unmarked.

Insegnanti capaci sono partite ...

Insegnanti capaci sono partiti ...

Recommendations 71 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R The value common is applied only to the present participle, where the gender is undecided in the lexicon, even though it can be disambiguated in the corpus.

Attribute value It. example It. tag

Gender masculine amato (past part) V/prm

(un) amante(pr part) V/ppm feminine amata (past part) V/prf

(un') amante (pr part) V/ppf neuter - - common (l') amante (pr part) V/ppn

10.2.6 Number This has only two values, singular and plural.

Attribute value It. example It. tag

Number singular amo, ami, ama V/s

plural amiamo, amate, amano V/p invariant - -

10.2.7 Combination of features Italian verbs have a maximum of 52 different inflected word-forms derived from different combinations of morphological features for simple morphological units (i.e. excluding compound forms). In summary, the following combinations of morphological features are applied, depending on Verb-form and, within a Verb-form, on different Moods.

SIMPLE FORMS Finiteness Verb-Form/Mood gramm. features

finite ind Tense Person Number

sub Tense Person Number impr Tense Person Number cond Tense Person Number non-finite inf Tense

ger Tense part Tense Number Gender

Compound forms have the following combinations:

COMPOUND FORMS Finiteness grammatical features

finite Tense Person Number Gender non-finite Tense Gender Number

Recommendations 72 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R Compound forms, because of the presence of the participle, have the feature Gender. 10.2.8 Enclitic Phenomenon In Italian, pronouns (pronominal particles) can accompany verbs in order to form: (i) the pronominal form of the verb (ii) the reflexive form (iii) the reciprocal form

As far as the lexicon is concerned, the possibility of taking clitics of different types (reflexive, reciprocal, etc.) has to be encoded in the entry, but depends on the syntactic type of the verb; these problems should therefore be dealt with on the syntactic level.

In these forms, the unstressed variant of the pronoun is, in general, separated graphically by the verb (it precedes the verb), except with the infinitive, the gerund and the imperative form and sometimes (rarely) with the past participle.

egli si lava lavarsi, lavandosi, lavati, lavatosi

At the corpus level, a clitic pronoun can be attached to the verb - when it expresses the direct object, dirlo (dire+lo = to say it) or the indirect object, dirgli (dire+gli = to say to him), - when it expresses a place adverb: andarci (andare+ci) (to go there).

Sequences of a verb with more than one pronoun in a unique graphical form are found if both the direct object and the indirect object of a pronominal verb form are represented by a clitic pronoun: ex. dandomelo (dando+me+lo) = giving it to me). These types of compounds can present the phenomenon of epenthesis, i.e. the insertion of a letter for euphonetic reasons: ex. dandoglielo (dando+gli+e (epenthetic)+lo) = giving it to him).

As to the dictionary encoding, `verb-pronoun' compounds have to be represented; the problem of encoding specific `verb-more-than-one-pronoun' compounds only concerns corpus encoding. The strategy adopted in tagging them is to assign differents tags to the different parts forming the word-token, with a special mark which maintains the graphical links, thus permitting the recovery of the unique graphical form.

Recommendations 73 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R 10.3 Application to German 10.3.1 Type The IMS-tagset considers the following 3 types of verbs:

ffl modals: m"ussen, k"onnen, d"urfen, wollen, sollen, m"ogen ffl auxiliaries: haben, sein, werden ffl full verbs: all other verbs

Attribute value example tag Verb-Type modal wollen VMINF

auxiliar haben VAINF full verb gehen VVINF

10.3.2 Finiteness, Verb-Form/Mood Finite and non-finite verb forms are not distinguished by a single feature in the IMS-tagset. Non-finite verb forms are infinitive, past participles. Present participle forms are classified as adjectives.

Example: das lachende/ADJA Kind

er kommt lachend/ADJD herein

Attribute value example tag Finite Verb-Form/Mood infinitive k"onnen VMINF

gehen VVINF infinitive with anzukommen VVINFZU incorporated "zu" past participle gewesen VAPPF

geliebt VVPPF

Non-Finite Verb-Form/Mood indicative (er) geht VVFIN:3.Sg.Pres.Ind

subjunctive (er) gehe VVFIN:3.Sg.Pres.Konj imperative geh! VVIMP:2.Pl

10.3.3 Tense The feature tense applies only to finite verb forms (but not to imperatives). The values of tense reflect only synthetic tensed verb forms, i.e. present tense (Pr"asens) and simple past (Pr"ateritum). Compound tenses (eg. ich habe gesehen) are not included.

Attribute value example tag

Tense present (ich) gehe VVFIN:1.Sg.Pres.Ind

past (ich) ging VVFIN:1.Sg.Past.Ind

Recommendations 74 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R 10.3.4 Person The feature person applies only to finite verb forms. In the IMS-Tagset it is also added to imperatives, where it would not be necessary since German imperatives exist only for 2nd person singular and plural.

Attribute value example tag

Person first (ich) gehe VVFIN:1.Sg.Pres.Ind

second (du) gehst VVFIN:2.Sg.Pres.Ind third (er) geht VVFIN:3.Sg.Pres.Ind

10.3.5 Number The feature number applies only to finite verb forms and imperatives.

Attribute value example tag

Number singular (ich) gehe VVFIN:1.Sg.Pres.Ind

geh! VVIMP:2.Sg.Pres.Ind plural (wir) gehen VVFIN:1.Pl.Pres.Ind

geht! VVIMP:2.Pl.Pres.Ind

10.3.6 Gender The feature gender does not apply to German verb forms. It occurs only with participles which are used as adjectives (and will be tagged as such).

Example: er hat es getan/VVPPF

nach getaner/ADJA Arbeit

Recommendations 75 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R 10.4 Application to English English verbs have no Gender distinction. In addition, the deployment of the Person, Number, Mood and Tense attributes is extremely limited. In the past tense, there are no distinctions of Person and Number, except for the solitary case of the verb to be, which has a singular form was distinct from the plural form were. In the present tense, except again for the verb to be, only two forms of each verb occur: the -s form in the third person singular, and the base form for all other combinations of person and number. These observations apply to the indicative mood. As for the other moods (subjunctive and imperative), the subjunctive is rarely used, and the imperative is invariable. The subjunctive is also invariable, except for a vestigial past subjunctive of the verb to be, in the singular use of were. Because of extensive syncretism in the English verb, the historical paradigms of Person, Number, Case, and Mood are barely sustainable in the description of modern English. The base form, if regarded as multi-functional, has at least the following morphological functions for all verbs except to be: (a) singular present tense indicative, 1st person (b) singular present tense indicative, 2nd person (c) plural present tense indicative (d) imperative (e) present tense subjunctive (f) infinitive Since it is impractical, however, given the current capabilities of tagging software, to resolve automatically the ambiguity of these six morphological functions, it is a common practice to assign a single value to the base form, or else to assign two values, one for the finite and one for the non-finite functions. Because of this, the tables below show two tagsets: one tagset representing the 6 attribute-values above, and a reduced tagset ("RTags"), which resembles most tagsets so far used for the English language in reducing the six values to two.

10.4.1 Type

Attribute values Examples Tags RTags Type Main eaten VVPp VVN

Auxiliary has (eaten) VP(Ind)PrS3 VPZ

The value "auxiliary verb" is not represented directly in the tagset, but is implied by the use of VP (for primary auxiliaries) and VM (for modal auxiliaries). Alongside the principal values of "Main" and "Auxiliary" verbs, it would have been useful to add a third, intermediate category of "semi-auxiliaries", for expressions such as be going to, have to, have got to, and be able to. However, for simplicity, this category has been omitted from the tagset.

10.4.2 Finiteness

Attribute values Examples Tags RTags Finiteness finite eats VV(Ind)PrS3 VVZ

nonfinite eating VVIng VVG

The values "finite" and "nonfinite" are not distinguished in the tagset, since they are predictable Recommendations 76 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R from other values, viz. those for "verb-form" and for "mood". 10.4.3 Verb-form Two attributes in this description of English, Verb-form and Mood, are derived from a single attribute (Verb-form/Mood) used elsewhere in this document. Verb-form applies only to nonfinite verbs, whereas Mood applies only to finite verbs:

Attribute values Examples Tags RTags Verb-form infinitive (to) eat VVInf VVI

ing-form eating VVIng VVG past part. eaten VVPp VVN

The value "ing-form" applies to all verb forms ending in the inflectional suffix -ing. In modern English, the distinction between present participle and gerund, representing two different functions of the -ing-form, is difficult to draw and of questionable validity. Consequently, tagsets for English generally treat the -ing form as a unitary category. This is a language-specific value for English.

10.4.4 Mood This applies only to finite verbs. The value "indicative" can remain implicit in the tagset, being the default or unmarked mood.

Attribute values Examples Tags RTags Mood indicative eats VV(Ind)PrS3 VVZ

subjunctive eat VVSubPr VVB imperative eat VVImp VVB

10.4.5 Tense

Attribute values Examples Tags RTags Tense present (they) eat VV(Ind)PrP VVB

past (they) ate VV(Ind)Pa VVD

10.4.6 Person

Attribute values Examples Tags RTags Person 1st person (I) am VP(Ind)PrS1 VPM

2nd person (you) are VP(Ind)PrS2 VPR 3rd person (she) is VP(Ind)PrS3 VPZ

10.4.7 Number

Attribute values Examples Tags RTags Number singular was (eaten) VP(Ind)PaS VPDZ

plural were (eaten) VP(Ind)PaP VPDR

Recommendations 77 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R 10.4.8 Auxiliary Type

Attribute values Examples Tags RTags Aux.Type Primary had (eaten) VPPa VPD

Modal would (eat) VM VM

The auxiliaries in English subdivide into the primary verbs be, have, and do, which can also function as main verbs, and the modal auxiliaries such as can, will, and would, which are uninflected, and always function as auxiliaries. These are language-specific values.

Recommendations 78 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R 10.5 Application to Dutch The thirteen Inflectional Verb features distinguished in the CELEX Wordforms lexical database are represented twice, once separately as a Yes or No code and once in combination, in a final composed `Flection Type' tag containing those features which received a Yes tag in four columns. For example: the word form with final Flection Type tag `te1s' received Yes four times, once for present tense (t), once for singular (e), once for first person (1) and once for separable word form (s).

10.5.1 Verb-Type CELEX distinguishes four types of verbs. The first three (full verbs, modals and auxiliaries) are the same as in the German application; the fourth is the group of impersonal verbs which have only one (impersonal) subject, `het'. Compare French: `il pleut' (pleuvoir).

Attribute value Example Tag Verb-Type full verb afwassen zelfst.

auxiliary hebben hulp. modal lijken koppel. impersonal regenen onpers.

Note that many verbs are tagged with double or triple tags because they can have two or three values.

10.5.2 Verb-Form The CELEX Wordforms Lexicon distinguishes between infinitive, present participle, past participle, finite verb forms and some other forms.

Attribute value Example Tag Verb-Form infinitive gaan i

pres participle gaand pt past participle gegaan pv finite (hij) gaat imperative ga g

See below for other participle word forms. Finite verb forms are not marked as such, but the 14 present and past tense Flection Type tags which are distinguished are all finite verb forms.

As the degree of `language-specificness' of the following five distinctions is not clear, they are not classified as such.

Recommendations 79 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R Participle word forms The CELEX Wordforms Lexicon distinguishes three Present participle word forms and seven Past participle word forms.

pt Present participle lopend ptE Present participle with suffix e lopende ptEm Pres.part. with suffix e , plural(de) lopenden pv Past participle gecoordineerd pvC Past participle Comparative gecoordineerder pvE Past participle with suffix e gecoordineerde pvEe Idem of irregular verbs, singular beschonkene pvEm Past participle with suffix e, plural gedekoreerden pvEs Past part. with suffix e, separated rond gezeilde pvs Past participle, separated rond gezeild

Only the Past and Present participle word forms `pt', `pv' and `pvs' are verbal word forms. However, they can also function as adjectives (and thus as adverbs) or nouns. The other Participle word forms are only used as adjectives, adverbs or nouns, so they can be covered by adjectival, adverbial or nominal tags. From the perspective of lemmatization, however, it might be important to stress the verbal aspect of participles, since, as long as participles are not yet `lexicalized' (accepted as real adjectives or nouns with a meaning of their own, differing from the meaning of the verb they are derived from) they do not have a lemma of their own, but are still to be lemmatized as words deriving from a verbal lemma.

The table below is a proposal for Dutch, not a CELEX table ! This table applies to the verb forms Infinitive, Present and Past participle.

Attribute value Example Tag Use (non-finite) verbal (hij heeft) geleerd PartVerb

adjectival (hij is) geleerd PartAdj nominal Leren (motivert) Part/Nom

(Het) geleerde

10.5.3 Mood The feature mood only applies to finite verb forms.

Attribute value Example Tag

Mood indicative (hij) gaat

subjunctive kome (wat komt) a

The Indicative is not marked as such, most of the finite verb forms being indicative. Only the few subjunctive verb forms have a special tag: `a'. In Dutch the subjunctive only occurs in the

Recommendations 80 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R third person singular. Phrases containing subjunctive verb forms mostly belong to the written language and possess an archaic or idiomatic quality. The subjunctive generally occurs in the Present tense. Only one Past tense form remains (ware).

10.5.4 Tense In the CELEX Wordforms Lexicon the feature Tense applies to finite verb forms and to participle verb forms. There are 10 different Present tense tags and 4 different Past tense tags. All of them are composed tags starting with `t' (Present time) or `v' (Past time). See also the list of Present and Past participle tags above.

Attribute value Example Tag

Tense present (ik) ga t

past (ik) ging v

10.5.5 Person The three different person-values are only distinguished in the Present tense singular. Present tense plural has only one verb form for the 1st, 2nd and 3rd person. The Past tense has one singular verb form and one plural verb form. The person-tags `1', `2' and `3' always appear in a composed Flection Type tag starting with: `te' (present tense singular).

Attribute value Example Tag

Person first (ik) ga te1

second (jij) gaat te2

ga (jij) te2I third (hij) gaat te3

See above the feature Inversion. 10.5.6 Number The number-tags `e' and `m' always appear in a composed Flection Type tag in combination with `t' (Present tense) or `v' (Past tense).

Attribute value Example Tag

Number singular (ik) ga te

(ik) ging ve plural (wij) gaan tm

(wij) gingen vmI

Recommendations 81 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R 10.5.7 Gender The feature gender does not apply to Dutch verb forms. It occurs only in Present and Past participles used as nouns or as adjectives. In these cases gender will have to be determined from the context (Attribute value context of the Gender table in nouns).

10.5.8 Main-Verb Function CELEX also distinguishes subcategorization values in full verbs (called `lexical' verbs). The three possible values are intransitive, transitive and reflexive.

Attribute value Example Tag Main-Verb Funct. intransit vallen intrans.

transit kopen trans. reflex vergissen wederk.

10.5.9 Auxiliary Another syntactical verb feature tagged in the Lemmas Lexicon is the type of auxiliary with which a verb is conjugated. There are three tags:

Attribute value Example Tag

Auxiliary hebben doen hebben

zijn groeien zijn hebben or zijn volgen hebben/zijn

Note that many verbs are tagged with double or triple tags because they can have two or three values. This is the common Dutch dictionary practice to mark verb use.

10.5.10 Other Verb-Form features In the CELEX Lemmas and Wordforms Lexica two further important verbal features have been tagged: separability and inversion. Another feature, word order , is distinguished in the other CELEX database mentioned above. In Dutch different word order brings about different word forms in the case of separable verbs.

10.5.11 Separability This is tagged in the Syntactical part of the Lemmas Lexicon with a Yes or No tag and in the Wordforms Lexicon with an `s' in the composed final Flection Type tag:

Recommendations 82 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R

Attribute value Example Tag Separability separable geef aan te1s

geeft aan te2s geeft aan te3s geven aan tms gaf aan ves gaven aan vms

Separable verbal word forms have a verbal and a non-verbal part and may occur at different places in the sentence. So another solution (in terms of corpus linguistics) would be to tag the verbal part as a separable verb form and the prepositional part as the non-verbal part of a separable verb, this last feature becoming a value of Adpositions.

10.5.12 Inversion This occurs in interrogative phrases and only brings about a different word form in the Present tense singular, second person. It is an inflectional feature tagged in the Wordforms Lexicon with an `I'.

Attribute value Example Tag

Inversion inverted ga (jij)? te2I

geef (jij) aan te2Is

te2I= present tense, singular, second person, inverted wordform te2Is= present tense, singular, second person, inverted and separable wordform

The inversion tag is important from the point of view of disambiguation, since the inverted form of the second person singular Present tense is the same as the non-inverted form of the first person singular Present tense!

10.5.13 Word order separable verbs This distinction comes from a CELEX database; the tag is proposed by the Dutch correspondent.

Attribute value Example Tag Word order - sep.verb main clause (hij) geeft aan ..MC

sub-clause (dat hij) aangeeft ..SC

Recommendations 83 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R Instances of separable verbs are written together in sub-clauses and written separately in main clauses!

10.5.14 Politeness Politeness is not represented in the CELEX tag sets. In Dutch polite verb forms correspond to the 2nd person singular and plural. The personal pronoun expressing politeness is: `U'.

Recommendations 84 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R 10.6 Application to Spanish 10.6.1 Verb-Form and Mood In Eurotra's terminology Verb-form corresponds to the attribute "e mstype" and its organization is as follows:

Attribute Value Example Tag e mstype finite cantamos finite

infinitive cantar infin gerund cantando gerund pastpart cantado pastpart

As for Mood ("e mood" in Eurotra) we have three possible values:

Attribute Value Example Tag e mood indicative canto indicative

subjunctive cantase subjunctive imperative canta imperative

It is not difficult to redistribute this organization so that it follows the strategy suggested in the present proposal:

Attribute value ex. tag Verb-Form finite - -

Attribute value example tag

Mood indicative canto

subjunctive cantase imperative canta Verb-Form non-finite - -

Attribute value example tag

Mood infinitive cantar

gerund cantando pastpart cantado

Recommendations 85 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R 10.6.2 Tense

Attribute Value Example Tag Tense present canto

past cant'e imperative canta future cantar'e conditional cantar'ia

A controversial point here concerns the interpretation of conditionals as tenses, in contrast to how they are usually considered in traditional grammars. This treatment is based on the assumption that the conditionals have a temporal value that correponds to a "future of the past".

"Compound tenses", as in Italian, are formed with an auxiliary. 10.6.3 Person

Attribute Value Example Tag Person first (yo) como

second (tu) comes third (el) come

10.6.4 Number

Attribute Value Example Tag Number singular (yo) como sing

plural (ellos) comen plu

10.6.5 Gender

Attribute Value Example Tag Gender feminine cansada fem

masculine cansado masc

Gender is only pertinent to past participle forms.

Recommendations 86 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R 10.7 Application to French (Corpus) 10.7.1 Type The IBMF tagset has special tags for the two auxiliaries, avoir and ^etre, otherwise similar (indication of person only) to the VERB tag used for "main" verbs. The distinction between the two auxiliaries would require a tagset-specific tag in EAGLES, or a special feature LU= indicating the exact lexical unit considered.

Attribute value Example Tag Type main lisait VERB3

auxiliary avait, 'etait AUXA3, AUXE3

10.7.2 Finiteness and Verb-Form/Mood As in the case of Italian, there is no specific encoding for the finiteness feature, which is selfcontained in the mood information. In French, the word forms allow for much of the disambiguation between moods, tenses, etc. When there is ambiguity (e.g. between present indicative and present subjunctive), the type of modelling involved with the IBMF tagset (i.e. stochastic tri-pos) would be unable to decide anyway. This is why the IBMF tagset does not distinguish all main moods/tenses, only Infinitive, Participle, and the rest (tag VERB).

Attribute value Example Tag Verb-Form indicative (je) lis VERB1

subjunctive " imperative " conditional " infinitive lire VINF participle lu PPASMS gerund - - supine - -

10.7.3 Tense Only in the case of participles is the present/past distinction made in the IBMF tagset. Assuming we already have set Fin=non-finite and VFM=part:

Attribute value Example Tag Tense present lisant PPRE

past lu PPASMS

10.7.4 Person In the IBMF tagset, the person is numbered 1 to 6, thus avoiding the need for number information. The coding of an IBMF verb tag into the EAGLES scheme would therefore imply using both the Person and Number attributes.

Recommendations 87 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R

Attribute value Example Tag Person 1 (je) vais, (nous) allons VERB1, VERB4

2 (tu) vas, (vous) allez VERB2, VERB5 3 (il) va, (ils) vont VERB3, VERB6

10.7.5 Gender In French, gender is only marked in the past participle.

Attribute value Example Tag Gender masculine ferm'es PPASMP

feminine ferm'ee PPASFS

10.7.6 Number As mentioned for the Person attribute, the coding of an IBMF verb tag into the EAGLES scheme implies combining Person and Number attributes in EAGLES. Otherwise, number is used only for Past participles in the IBMF tagset.

Attribute value Example Tag Number singular ferm'e PPASMS

plural ferm'es PPASMP

10.7.7 EAGLES features not applicable Separability does not apply to French. Voice and Reflexivity are not marked as such in the verb form (they are built by adjunction of auxiliary or pronoun). Auxiliary would apply in French since all verbs build their compound forms with one of the two auxiliaries already mentioned. These and other features (Main-verb fuction, auxiliary function) are not coded in the IBMF tagset.

10.7.8 IBMF Tagset features not applicable in EAGLES The specification of the exact lexical unit for the auxiliary (tags AUXA AUXE, see above).

Recommendations 88 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R 10.8 Application to French (Lexicon) 10.8.1 Type ------------ ----------- ----------- ---- Attribute Value Example Code ------------ ----------- ----------- ---- Type main partir m

auxiliary avoir a ------------ ----------- ----------- ----

10.8.2 Finiteness Finiteness is redundant with mood in French, as shown in the table below: ------------ ----------- Verb Form/M. Finiteness ------------ ----------- indicative finite subjunctive finite imperative finite conditional finite infinitive non-finite participle non-finite ------------ -----------

We therefore decided not to encode it.

10.8.3 Verb Form/Mood ------------ ----------- ----------- ---- Attribute Value Example Code ------------ ----------- ----------- ---- Verb Form/M. indicative viens i

subjunctive vienne s imperative viens m conditional viendrais c infinitive venir n participle venu p ------------ ----------- ----------- ----

10.8.4 Tense ------------ ----------- ----------- ---- Attribute Value Example Code ------------ ----------- ----------- ---- Tense present viens p

imperfect venais i future viendrai f

Recommendations 89 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R

past vins s ------------ ----------- ----------- ----

10.8.5 Person ------------ ----------- ----------- ---- Attribute Value Example Code ------------ ----------- ----------- ---- Person first suis 1

second es 2 third est 3 ------------ ----------- ----------- ----

10.8.6 Number ------------ ----------- ----------- ---- Attribute Value Example Code ------------ ----------- ----------- ---- Number singular viens s

plural venons p ------------ ----------- ----------- ----

10.8.7 Gender ------------ ----------- ----------- ---- Attribute Value Example Code ------------ ----------- ----------- ---- Gender masculine venu m

feminine venue f ------------ ----------- ----------- ----

10.8.8 Combinations --------- ------------------------ Tag Example --------- ------------------------ V1-s-ip-m viens V2-s-ip-m viens V3-s-ip-m vient V1-p-ip-m venons V2-p-ip-m venez V3-p-ip-m viennent V1-s-ii-m venais V2-s-ii-m venais V3-s-ii-m venait V1-p-ii-m venions V2-p-ii-m veniez V3-p-ii-m venaient

Recommendations 90 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R V1-s-if-m viendrai V2-s-if-m viendras V3-s-if-m viendra V1-p-if-m viendrons V2-p-if-m viendrez V3-p-if-m viendront V1-s-is-m vins V2-s-is-m vins V3-s-is-m vint V1-p-is-m vinmes V2-p-is-m vintes V3-p-is-m vinrent V1-s-ip-a suis, ai V2-s-ip-a es, as V3-s-ip-a est, a V1-p-ip-a sommes, avons V2-p-ip-a etes, avez V3-p-ip-a sont, ont V1-s-ii-a etais, avais V2-s-ii-a etais, avais V3-s-ii-a etait, avait V1-p-ii-a etions, avions V2-p-ii-a etiez, aviez V3-p-ii-a etaient, avaient V1-s-if-a serai, aurai V2-s-if-a seras, auras V3-s-if-a sera, aura V1-p-if-a serons, aurons V2-p-if-a serez, aurez V3-p-if-a seront, auront V1-s-is-a fus, eus V2-s-is-a fus, eus V3-s-is-a fut, eut V1-p-is-a fumes, eumes V2-p-is-a futes, eutes V3-p-is-a furent, eurent V1-s-sp-m finisse V2-s-sp-m finisse V3-s-sp-m finisse V1-p-sp-m finissions V2-p-sp-m finissiez V3-p-sp-m finissent V1-s-si-m finisse V2-s-si-m finisse V3-s-si-m finit V1-p-si-m finissions V2-p-si-m finissiez

Recommendations 91 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R V3-p-si-m finissent V1-s-sp-a sois, aie V2-s-sp-a sois, aies V3-s-sp-a soit, ait V1-p-sp-a soyons, ayons V2-p-sp-a soyez, ayez V3-p-sp-a soient, aient V1-s-si-a fusse, eusse V2-s-si-a fusses, eusses V3-s-si-a fut, eut V1-p-si-a fussions, eussions V2-p-si-a fussiez, eussiez V3-p-si-a fussent, eussent V2-s-mp-m viens V1-p-mp-m venons V2-p-mp-m venez V2-s-mp-a sois, aie V1-p-mp-a soyons, ayons V2-p-mp-a soyez, ayez V1-s-cp-m viendrais V2-s-cp-m viendrais V3-s-cp-m viendrait V1-p-cp-m viendrions V2-p-cp-m viendriez V3-p-cp-m viendraient V1-s-cp-a serais, aurais V2-s-cp-a serais, aurais V3-s-cp-a serait, aurait V1-p-cp-a serions, aurions V2-p-cp-a seriez, auriez V3-p-cp-a seraient, auraient V----n--m venir V----n--a etre, avoir V-ms-ps-m venu V-fs-ps-m venue V-mp-ps-m venus V-fp-ps-m venues V-ms-ps-a eu V-fs-ps-a eue V-mp-ps-a eus V-fp-ps-a eues V----pp-m venant V----pp-a etant, ayant V----ps-m semble' V----ps-a ete' --------- ------------------------

Recommendations 92 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R Note: We have decided to encode the past participle of the auxiliary verb ^etre, the copulative verbs (sembler etc.) and the impersonal verbs (e.g. falloir), with the "not applicable" feature (-) instead of the "neutral" or "masculine" one. The notion of neutrality, though used for pronouns in the present model, is not a very traditional notion in French grammars and in any case the neutral feature does not bring any further information than the not applicable one. As for the masculine feature, syntactic considerations led us to drop it in order to maintain consistent agreement marks inside a sentence.

Recommendations 93 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R 10.9 Application to Portuguese The following set of morphological features are pertinent to Verbs in the Portuguese model GENELEX: Mood, Tense, Person, Gender, Number. The generation of the inflected forms of a given verb system is based on the various combinations of these features. Some features are not pertinent to the characterization of some wordforms, as we will try to explain below.

10.9.1 Type The value auxiliar is not encoded in the morphological layer of the Portuguese model of GENELEX, but it is required on its syntactic layer, where it is applied to the verbs ter and haver. The value auxiliar is therefore pertinent to the Portuguese language.

10.9.2 Finiteness This feature can be unproblematically inferred from the feature Mood, as the schema below clearly shows. The redundant application of these two features to Portuguese could be represented as follows:

---------------------- ----------------------- Verb Form/Mood Finiteness ---------------------- ----------------------- indicative finite subjunctive finite imperative finite conditional finite infinitive non-finite participle non-finite ---------------------- -----------------------

The feature Finiteness is therefore not required by the Portuguese language.

10.9.3 Mood From the values proposed for Mood, the following values are used in the Portuguese application:

------------- --------------- ------------- ---------- Attribute Value Example Tag ------------- --------------- ------------- ---------- Mood indicative como

subjunctive coma imperative come conditional comeria infinitive comer participle comido supine ------------- --------------- ------------- ----------

Recommendations 94 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R 10.9.4 Tense Given that Plus-quam-Perfect is also a simple tense in the Portuguese language, we stress the need for integrating a specific value for this tense, so that the relevant simple wordforms can be handled in morphology.

------------- ------------------------ -------------- ------------ Attribute Value Example Tag ------------- ------------------------ -------------- ------------ Tense present como

imperfect comia perfect comi future comerei past comido ============= ======================== ============== ============ l.spec plus.quam.perfect comera ------------- ------------------------ -------------- ------------

10.9.5 Person The values proposed in the tables are pertinent to Portuguese. However, in order to cope with typical aspects of Politeness reflected in verbal inflection, the GENELEX model built for the Portuguese language made a particular use of this attribute - instead of a feature person, a decomposition of this is used: person-deixis and person-conc. Since in Portuguese polite verb forms correspond to the 3rd person, these two features allow an enhanced specification of verbal wordforms where the value `person=3' could give rise to ambiguity (remark that ambiguity may be not solved by context, because in Portuguese you can drop the np subject.) In the GENELEX application such verbal wordforms (e.g. come (`eats' or `eat')) are thus marked in two ways: polite verbal wordforms (as in `voce'2 come' (`you eat')) bear the combination of values `person-deixis=2' and `person-conc=3'; true `person=3' wordforms (as in `ele come' (`he eats')) bear the combination of values `person-deixis=3' and `person-conc=3'.

10.9.6 Gender This feature is pertinent only to the Mood participle.

----------- --------------- --------------- ------------------ Attribute Value Example Tag ----------- --------------- --------------- ------------------ Gender masculine comido

feminine comida neuter common ----------- --------------- --------------- ------------------

Only the values masculine and feminine are used.

Recommendations 95 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R 10.9.7 Number ------------- ------------- ------------ --------------------- Attribute Value Example Tag ------------- ------------- ------------ --------------------- Number singular como

plural comemos invariant ------------- ------------- ------------ ---------------------

This has only two values, singular and plural.

Recommendations 96 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R 10.9.8 Combination of features In the GENELEX application, Portuguese verbs have a maximum of 91 different inflected wordforms derived from different combinations of morphological features for simple morphological units (i.e. excluding compound forms). They are the following:

----------------- ---------------------------

Example ----------------- --------------------------- 01 Ind-Pr-1-1---S como 02 Ind-Pr-2-2---S comes 03 Ind-Pr-2-3---S come 04 Ind-Pr-3-3---S come 05 Ind-Pr-1-1---P comemos 06 Ind-Pr-2-2---P comeis 07 Ind-Pr-2-3---P comem 08 Ind-Pr-3-3---P comem 09 Ind-Im-1-1---S comia 10 Ind-Im-2-2---S comias 11 Ind-Im-2-3---S comia 12 Ind-Im-3-3---S comia 13 Ind-Im-1-1---P comi'1amos 14 Ind-Im-2-2---P comi'1eis 15 Ind-Im-2-3---P comiam 16 Ind-Im-3-3---P comiam 17 Ind-Fu-1-1---S comerei 18 Ind-Fu-2-2---S comera'1s 19 Ind-Fu-2-3---S comera'1 20 Ind-Fu-3-3---S comera'1 21 Ind-Fu-1-1---P comeremos 22 Ind-Fu-2-2---P comereis 23 Ind-Fu-2-3---P comera'3o 24 Ind-Fu-3-3---P comera'3o 25 Ind-Pa-1-1---S comi 26 Ind-Pa-2-2---S comeste 27 Ind-Pa-2-3---S comeu 28 Ind-Pa-3-3---S comeu 29 Ind-Pa-1-1---P comemos 30 Ind-Pa-2-2---P comestes 31 Ind-Pa-2-3---P comeram 32 Ind-Pa-3-3---P comeram 33 Ind-Pq-1-1---S comera 34 Ind-Pq-2-2---S comeras 35 Ind-Pq-2-3---S comera 36 Ind-Pq-3-3---S comera 37 Ind-Pq-1-1---P come'2ramos

Recommendations 97 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R 38 Ind-Pq-2-2---P come'2reis 39 Ind-Pq-2-3---P comeram 40 Ind-Pq-3-3---P comeram 41 Con-Pr-1-1---S comeria 42 Con-Pr-2-2---S comerias 43 Con-Pr-2-3---S comeria 44 Con-Pr-3-3---S comeria 45 Con-Pr-1-1---P comeri'1amos 46 Con-Pr-2-2---P comeri'1eis 47 Con-Pr-2-3---P comeriam 48 Con-Pr-3-3---P comeriam 49 Sub-Pr-1-1---S coma 50 Sub-Pr-2-2---S comas 51 Sub-Pr-2-3---S coma 52 Sub-Pr-3-3---S coma 53 Sub-Pr-1-1---P comamos 54 Sub-Pr-2-2---P comais 55 Sub-Pr-2-3---P comam 56 Sub-Pr-3-3---P comam 57 Sub-Im-1-1---S comesse 58 Sub-Im-2-2---S comesses 59 Sub-Im-2-3---S comesse 60 Sub-Im-3-3---S comesse 61 Sub-Im-1-1---P come'2ssemos 62 Sub-Im-2-2---P come'2sseis 63 Sub-Im-2-3---P comessem 64 Sub-Im-3-3---P comessem 65 Sub-Fu-1-1---S comer 66 Sub-Fu-2-2---S comeres 67 Sub-Fu-2-3---S comer 68 Sub-Fu-3-3---S comer 69 Sub-Fu-1-1---P comermos 70 Sub-Fu-2-2---P comerdes 71 Sub-Fu-2-3---P comerem 72 Sub-Fu-3-3---P comerem 73 Imp----2-2---S come 74 Imp----2-3---S coma 75 Imp----1-1---P comamos 76 Imp----2-2---P comei 77 Imp----2-3---P comam 78 Inf----1-1---S comer 79 Inf----2-2---S comeres 80 Inf----2-3---S comer 81 Inf----3-3---S comer 82 Inf----1-1---P comermos 83 Inf----2-2---P comerdes 84 Inf----2-3---P comerem

Recommendations 98 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R 85 Inf----3-3---P comerem 86 Inf----------- comer 87 Par-Ps-----M-S comido 88 Par-Ps-----F-S comida 89 Par-Ps-----M-P comidos 90 Par-Ps-----F-P comidas 91 Ger----------- comendo

We stress that there is no need to define any specification labelling the typically Portuguese `inflected infinitive'. In fact, it is suitably handled by combining `mood=infinitive' with the values of attributes `number' and `person' (the last is decomposed in the Portuguese GENELEX model, as we explained under the subsection `Person' above).

10.9.9 Clitics This feature is pertinent to Portuguese verbs, but the assignment of a feature concerning clitics to verbs depends on the syntactic classification of the verb. For the time being, verbs are not yet encoded wrt clitics in the Portuguese GENELEX morphological layer.

Recommendations 99 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R 10.10 Application to Danish In Danish, the verbal system comprises a very limited number of inflectional forms. There is no morphological distinction for person and number; moods and tenses are limited as well. Thus, the set of morphological features pertinent to Danish verbs are: Type, Tense, Mood (distinction between indicative and imperative only) and Voice (distinction of active and -s passive).

10.10.1 Type

Attribute value Example Tag

Type Main skriver vb mainv

Auxiliary har (skrevet) vb aux Passive auxiliary bliver (skrevet) vb auxpass Modal skulle (skrive) vb modv

The type `passive auxiliary' can be regarded as a language-specific feature; in Danish passive forms can be generated by means of the morphological suffix `-(e)s' or by means of the passive auxiliaries `blive' and `vaere', respectively. This feature is used within the EDEMD.

10.10.2 Finiteness In general, the finiteness is implicitly given by the particular tag of the verb-form occuring in the corpus, e.g. `skriver' is indicative present active which is a finite form. Thus, in corpus tagging the distinction finite/non-finite is not needed. However, in the Eurotra description of Danish finite and non-finite forms have been distinguished.

Attribute value Example Tag Finiteness finite skriver vb pres act

non-finite skrivende vb pres ptc

10.10.3 Verb-form Here we follow the tagging for English: the attribute Verb-form applies to non-finite verb forms, Mood applies to finite verb forms only.

Attribute value Example Tag Verb-form infinitive skrive vb inf

perfect participle skrevet vb pf.ptc present participle skrivende vb pres.ptc

A perfect participle remains uninflected when used in verbal function; in adjectival (attributive) function perfect participle forms are inflected like adjectives (without inflectional comparison).

Recommendations 100 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R 10.10.4 Mood In corpus tagging the mood `indicative' will be left unmarked, because this is the most frequently used - and thus the default - mood in texts. In corpora a few archaic examples of other moods do occur (e.g. subjunctive) but they are not taken into account here because of their very low frequency.

Attribute value Example Tag

Mood indicative (vi) skriver vb pres ind

imperative skriv vb imp

10.10.5 Tense Compound tenses (e.g. auxiliary + past participle for past tenses) are not included.

Attribute value Example Tag

Tense present (de) skriver vb pres

past (du) skrev vb past

10.10.6 Number In modern Danish, there is no difference between singular and plural verb-forms.

10.10.7 Person In modern Danish, there is full syncretism within the person paradigm of verb-forms, i.e. one single morphological form covers all Person + Number combinations of a given tense/mood. Therefore, the attributes Number and Person are not applied to Danish verbs. In a corpus a few archaic forms of the plural may occur.

10.10.8 Gender Danish verbs have no Gender distinction.

10.10.9 language-specific feature: Voice In Danish, there is a morphological difference between active and passive forms in the present and the past. (The inflectional ending of the passive is always -(e)s.) Passive forms can also be composed by means of an auxiliary (auxpass) + perfect participle. Perfect participle forms of a few verbs may also occur as -(e)s passive forms.

Recommendations 101 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R

Attribute value Example Tag

Voice active skriver vb pres act

-s passive skrives vb pres s-pas

Recommendations 102 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R 10.11 Application to Greek Potentially, Greek verbal entries are characterised for the following features: Finiteness, Verbform/Mood, Tense, Person, Number, Gender, Aspect, Voice.

Depending on the combination of these features, all inflected forms of the Greek verbal system can be generated. The combination of the features is dependent upon constraints, resulting, in certain cases, in the total absence of certain features or the exclusion of certain values. More information on these constraints is given in the following sections.

The Greek verbal system presents both simple and compound forms. Compound forms are the result of two different combinations:

- two verbs, the auxiliary and the main verb (the main verb is in the "infinitival" or past participial form) for the formation of certain tenses,

- a particle and the verb, for the formation of the subjunctive mood and the conditional tenses. In the following tables, compound forms and the appropriate values and tags they should be assigned are put in parentheses, given that their recognition requires the use of a multi-word tagger.

10.11.1 Type This feature is not currently coded in the Greek Morphological Lexicon. However, it can be applied to the Greek language. In this case, the table of the attribute-value set could take the following form:

Attribute value Gr. example Gr. tag

Type main ghrafw aux ehw cop eimai l-spec impers prepei The verb "eimai" in Greek acts as a main, copular or auxiliary verb. In the majority of the cases, the linguistic context contributes to the resolution of the ambiguity:

To vivlio einai panw sto trapezi (prep. - main) To vivlio einai teleiwmeno (past. part. - aux)

To vivlio einai katharo (adj. - cop.)

The verb "ehw" acts both as an auxiliary and as a main verb. Again, the linguistic context serves for disambiguation purposes:

Recommendations 103 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R

O Ghiannys ehei spiti (noun - main) O Ghiannys ehei teleiwsei (inf. - aux)

Impersonal verbs are peculiar to the Greek language. Under this class, we include verbs used only in the third singular person, which do not take a subject. In the case of verbs denoting weather phenomena, there is no subject, while in the case of impersonal verbs denoting modality, they are followed by a secondary clause :

Exw vrehei O Kwstas prepei na fughei O Kwstas kai y Ghianna prepei na fughoun

10.11.2 Finiteness and Verb-Form/Mood At its previous form, the Greek Morphological Lexicon did not code the feature Finiteness, while the features Verb-Form and Mood were kept distinct, the first one taking as values finite and participle, and the latter indicative and imperative. However, a combination of these two features could give information on finiteness. Apart from participles, all other simple forms in Greek are finite and may take the values indicative and imperative as regards mood.

In the current re-organisation of the lexicon, it has been decided to abandon this scheme in favour of the one proposed by EAGLES. The following table shows the new categorisation.

(For the formation of the subjunctive and the conditional mood a compound form is used, as presented in the following table.)

Attribute value Gr. ex. Gr. tag Finiteness non-finite ghrafontas VbNf

Attribute value Gr. example Gr. tag Verb-Form/Mood infinitive eisahthei VbNfIf

participle eisaghomenos VbNfPp

Finiteness finite eisaghw VbFi

Attribute value Gr. example Gr. tag Verb-Form/Mood indicative eisaghw VbFiId

imperative ghrapse VbFiMp (subjunctive) (na ghrapsw) (VbFiSj)

In the new form of the ILSP Lexicon, we have also decided to recognise as a distinct entity the "infinitival" form in Greek. This form is only used for the generation of certain compound tenses and cannot be found on its own; this form is the same morphologically with the third singular person used for the formation of the future tense or the subjunctive mood (again formed periphrastically, with the combination of a particle), and, thus, in order to avoid generation of unnecessary ambiguity, the value indicative was used for both forms in the former version of the Lexicon:

O Ghiannys tha teleiwsei (future - untns 3rd sing.)

O Ghiannys ehei teleiwsei (pres.perf. - inf.)

Recommendations 104 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R 10.11.3 Tense The tense system in Greek is traditionally perceived as similar to the one described for Romance languages. However, in the current codification system we have adopted a system more similar to the English one. Thus, the system is based on a combination of the Tense and Aspect features (for Aspect, see relevant section). Further constraints are applied on this attribute by the value of the mood. The tense system in Greek includes both simple and compound forms.

Attribute value Gr. example Gr. tag

Tense pres ghrafw VbFiIdPr

past eghrapsa VbFiIdPa (fut) (tha ghrapsw) (VbFiIdFu) l-spec untns ghrapsei VbNfIfUn

The value untns, specific to Greek, is used for the codification in the Lexicon of the "infinitival" form of the verb, form which, as already presented, never appears on its own but is used for the formation of compound tenses. It is also used for the form that combined with the appropriate particle forms the future tense and the subjunctive mood.

10.11.4 Person

Attribute value Gr. example Gr. tag

Person 1 ghrafw VbFiIdPr01

2 ghrafeis VbFiIdPr02 3 ghrafei VbFiIdPr03

For the codification of the infinitival form, the value of Person is left unspecified - i.e. the respective slot of the tag is given the value "00", which signifies that this particular feature does not apply to this wordform -, given that the person of the compound form is provided by the auxiliary verb:

O Ghiannys ehei fughei. - 3rd person

Ehw fughei. - 1st person

The value of person is left unspecified for the participial forms as well. Impersonal verbs are given the value 03 for the attribute of person, given that they are in fact morphologically marked as the third singular person.

Recommendations 105 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R 10.11.5 Number It has only two values, singular and plural.

Attribute value Gr. example Gr. tag

Number singular ghrafw VbFiIdPr01Sg

plural ghrafoume VbFiIdPr01Pl

As for Person, the attribute of Number is left unspecified for the "infinitival" form, because the number of the compound form is marked on the auxiliary verb:

O Ghiannys ehei fughei. - singular

Oloi ehoun fughei. - plural

In the case of active participial forms ending in -ontas, which function as adverbs, the value of number is also left unspecified:

Feughontas o Kwstas, ekleise tyn porta. - sing. Feughontas o Kwstas kai y Maria, mas apohairetysan. - plural

10.11.6 Gender and Case In the Modern Greek linguistic system, only two participial forms are recognised: * the active present participle, ending in -ontas, which functions as adverbs, and the passive participle of the present perfect tense, ending in -menos, -meny, -meno, which is considered to be a verbal adjective.

In addition to these two forms, a new form has been added in the recent years to the linguistic system of Modern Greek (probably, as a result of foreign languages' influence), namely:

* the active participle of the present perfect tense, formed periphrastically by the active present participle of the auxiliary verb "ehontas" and the infinitival form of the verb. Furthermore, there remain in Modern Greek certain participial forms from the Ancient Greek linguistic system. These forms, which were all verbal-adjectival types, are the following:

* the participle of the active present tense, ending in -wn, -ousa, -on, or -wn, - wsa, -wn, or -wn, -ousa, -oun, the participle of the passive present tense, ending in -omenos, -omeny, -omeno, the participle of the active past perfective tense, ending in -as, -asa, -an, the participle of the active past perfective tense, ending in -wn, - ousa, -on, and the participle of the passive past perfective tense, ending in -eis, -eisa, -en.

These forms have gradually lost their link to the verb they have been derived from and have come to be considered, in the majority of the cases, adjectives and, sometimes, nouns, depending on their function. However, very recently, a tendency to re-use them as verbal types and even to produce new participles of these forms, is observed in certain texts, mainly of the administrative type. Thus, in most of the modern theoretical grammars, these participial forms

Recommendations 106 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R are included under the categories of adjectives and nouns, although they recognise their origins from the verbal system.

At the present form of the ILSP lexicon, we have kept this distinction, i.e. the only types that are considered "participles" per se are the first two, i.e. the one of the active present tense and the one of the passive past perfect. This distinction is, however, under reviewing; still, a future category change of participles does not affect the status of the current proposal for a morpho-syntactic description of the linguistic system of Greek, given that the values of the present scheme are sufficient to describe them in either way.

The features of Gender and Case are used for the characterisation of passive past perfective participles, which are adjectival types of the verb. As adjectives, they agree in gender, number and case with the noun they modify. Therefore, apart from the values masc-fem and common, which do not aply here, all values are those presented in the section of nouns, and shown in the following table, which shows also the values that other participial forms would take if they were included under this category.

For all other verbal forms (finite forms, infinitive and active present participle), these features are left unspecified.

Attribute value Gr. example Gr. tag

Gender masculine eisaghmenos VbNfPpPa00SgMa

(eisaghwn) (VbNfPpPr00SgMa) (eisaghomenos) (VbNfPpPr00SgMa) (energhysas) (VbNfPpPa00SgMa) (eisahtheis) (VbNfPpPa00SgMa) (pathwn) (VbNfPpPa00SgMa) feminine eisaghmeny VbNfPpPa00SgFe

(eisaghousa) (VbNfPpPr00SgFe) (eisaghomeny) (VbNfPpPr00SgFe) (energhysasa) (VbNfPpPa00SgFe) (eisahtheisa) (VbNfPpPa00SgFe) (pathousa) (VbNfPpPa00SgFe) neuter eisaghmeno VbNfPpPa00SgNe

(eisaghon) (VbNfPpPr00SgNe) (eisaghomeno) (VbNfPpPr00SgNe) (energhysan) (VbNfPpPa00SgNe) (eisahthen) (VbNfPpPa00SgNe) (pathon) (VbNfPpPa00SgNe)

Recommendations 107 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R

Attribute value Gr. example Gr. tag

Case nom eisaghmenos VbNfPpPa00SgMaNm

gen eisaghmenou VbNfPpPa00SgMaGe acc eisaghmeno VbNfPpPa00SgMaAc

l-spec voc eisaghmene VbNfPpPa00SgMaVo

10.11.7 Aspect This attribute is used in combination with the attributes of Tense and Mood for the appropriate and unique characterisation of tense of all forms (simple and compound) in Greek.

Attribute value Gr. example Gr. tag

Aspect imperfective eghrafa VbFiIdPa01SgIp

perfective eghrapsa VbFiIdPa01SgPe (perfect) (ehw ghrapsei) (VbFiIdPa01SgPf)

The value perfect is not used in the Morphological Lexicon. It is supplied for future use of the annotation system, in order to appropriately distinguish the various simple and compound forms in the Greek tense system.

Below, we give a table showing the formation of various verbal finite forms in all combinations of mood, tense and aspect, only for the active voice. Compound forms and the tags they should take, if recognised by a multi-word tagger, are put inside parentheses.

Recommendations 108 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R

Attribute Attribute Attribute Gr. example Gr. tag

Mood Tense Aspect

ind pres imperfective ghrafw VbFiIdPr01SgIp ind past imperfective eghrafa VbFiIdPa01SgIp ind past perfective eghrapsa VbFiIdPa01SgPe ind (fut) imperfective (tha ghrafw) (VbFiIdFu01SgIp ) ind (fut) perfective (tha ghrapsw) (VbFiIdFu01SgPe) ind pres (perfect) (ehw ghrapsei) (VbFiIdPr01SgPf) ind past (perfect) (eiha ghrapsei) (VbFiIdPa01SgPf) ind fut (perfect) (tha ehw ghrapsei) (VbFiIdFu01SgP f) imp pres imperfective ghrafe VbFiMpPr02SgIp imp past perfective ghrapse VbFiMpPa02SgPe (subj) pres imperfective (na ghrafw) (VbFiSjPr01SgI p) (subj) past perfective (na ghrapsw) (VbFiSjPa01SgPe ) (subj) pres (perfect) (na ehw ghrapsei) (VbFiSjPr01SgPf) (cond) past imperfective (tha eghrafa) (VbFiCnPa01SgIp) (cond) past perfective (tha eghrapsa) (VbFiCnPa01SgPe) (cond) past (perfect) (tha eiha ghrapsei) (VbFiCnPa01SgPf)

10.11.8 Voice The Greek language recognises two values for the attribute of voice, the same proposed by EAGLES at Level 2a.

Attribute value Gr. example Gr. tag

Voice active ghrafw VbFiIdPr01SgImAv

passive ghrafomai VbFiIdPa01PlPePv

Recommendations 109 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R 10.12 Application to Swedish The Swedish verbs involve the following features; Type, Tense, Form/Mood, Gender/Number and Voice. We make no distinction in Person in modern Swedish.

Type Form/Mood Tense Voice Gender/Number normal imperative

present active aux.t finite

nonneuter aux.p supine

past passive neuter aux.m infinitive

plural participle

10.12.1 Type This is a lexical feature, although several verbs of the aux categories have polysemic variants functioning as normal verbs. The values pertinent to Swedish are normal, auxilliary-tense, auxilliary-passive, auxilliary-modal. (The term main verb is reserved for purely syntactic functions.)

Attribute value Example Tag Type normal skriver vb-mainv

auxilliary-tense har (skrivit) vb-auxt auxilliary-passive blir (skriven) vb-auxp auxiliary-modal skulle(skrivas) vb-auxm

10.12.2 Form/Mood This feature reflects the fact that in Swedish only imperative and indicative moods exist, of which only the latter is capable of morphosyntactic variation. (Certain remaining subjunctive forms, like vore, past subjunctive of vara 'to be', are best treated as exceptions.) The values are imperative, finite, supine, infinitive and participle. Of these, the values imperative and finite represents the finite forms and the rest the infinite; it is thus seen that a feature Finiteness is redundant for Swedish. The supine is used only as a main verb in composite tenses while the infinitive often functions as a kind of noun and the participles as adjectives.

Attribute value Example Tag Mood indicative skriver vb-pres-ind

imperative skriv vb-imp Form infinitive skriva vb-inf

supine skrivit vb-sup perfect participle skriven vb-pf.ptc present participle skrivande vb-pres.ptc

Recommendations 110 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R 10.12.3 Tense This feature is only applicable for finite forms and participles. The only morphosyntactically relevant values in Swedish are present and past.

Attribute value Example Tag Tense present skriver vb-pres

past skrev vb-past

10.12.4 Voice This is a language specific morphological feature, since Swedish, like other Scandinavian languages, possesses a special passive suffix (-s). (There also exists a composite or periphrastic passive, cf below.) The feature takes the values active and passive and is applicable to finite, supine and infinite forms.

Attribute value Example Tag Voice aktiv skrev vb-past-act

passiv skrevs vb-past-s-pas

NB: The feature Tense and Voice are realized in just that order; thus

vb(finite) past pass k"or -de -s

10.12.5 Gender/Number This feature is only applicable to the past participle, when used as the main verb of a composite passive. In that function it agrees morphologically with the subject in gender and number. Since the gender distinction is irrelevant in the plural, two features can be combined into one, taking the values nonneutral, neutral (implying singular) and plural. (In finite forms, Number is obselete in Swedish, having been dropped even in formal written language fifty years ago).

10.12.6 Person We make no difference in verbforms for Person in modern Swedish.

Recommendations 111 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R 10.13 Application to Catalan The following set of morphological features are pertinent to Verbs: Mood, Tense, Person, Gender, Number. These features, differently combined, dynamically generate the inflected forms of the Catalan verbal system, but not all the features and sometimes not all the values of a feature are applied to an inflected form.

10.13.1 Type Verbs like haver are marked as auxiliaries (in modern Catalan, esser is not an auxiliar verb), but in the corpus we don't disambiguate the cases of auxiliarity from the (rare) cases of intransitive uses.

10.13.2 Finiteness and Verb-Form/Mood In the Catalan corpus and lexicon there are no specific tags for the Feature Finiteness. As in the case of other Romance languages, it is implicitly contained in traditional division beetween non-finite forms of the verb (infinitive, participle, gerund) and the finite forms (which have four values for Mood): indicative, subjunctive, conditional, imperative.

----------------------------------------------------------------

Attribute value Ca. example Ca. tag ----------------------------------------------------------------

Finiteness non-finite - - ----------------------------------------------------------------

Attribute value Ca. example Ca. tag ----------------------------------------------------------------

Ver-F/Mood infinitive cantar V/if ----------------------------------------------------------------

participle cantat V/rms ----------------------------------------------------------------

gerund cantant V/ge ----------------------------------------------------------------

Finiteness finite - - ----------------------------------------------------------------

Attribute value Ca. example Ca. tag ----------------------------------------------------------------

Ver-F/Mood indicative canto V/1pi ----------------------------------------------------------------

subjunctive canti V/1ps ----------------------------------------------------------------

imperative canta V/2im ----------------------------------------------------------------

conditional cantaria V/1co ----------------------------------------------------------------

Recommendations 112 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R 10.13.3 Tense The four values of the feature Tense are only extensively applicable to the finite forms indicative and subjunctive of the verb, if we take into account just the simple inflected forms (and not the compound forms) of the verb (the value past in the case of past participles is completely redundant, because present participles can not be considered as verbal forms in modern Catalan).

--------------------------------------------------------------

Verb-Form/Mood value --------------------------------------------------------------

indicative --------------------------------------------------------------

Tense Ca. example Ca. tag --------------------------------------------------------------

pres canto V/1pi --------------------------------------------------------------

impf cantava V/1ii --------------------------------------------------------------

fut cantare V/1fu --------------------------------------------------------------

past canti V/1pt --------------------------------------------------------------

--------------------------------------------------------------

Verb-Form/Mood value --------------------------------------------------------------

subjunctive --------------------------------------------------------------

Tense Ca. example Ca. tag --------------------------------------------------------------

pres canti V/1ps --------------------------------------------------------------

impf cantes V/1is --------------------------------------------------------------

fut - - --------------------------------------------------------------

past - - --------------------------------------------------------------

10.13.4 Person Person is tagged by means of a number 1, 2, 3 (for singular); 4, 5, 6 (for plural) in the first position of the code of the inflected form.

Recommendations 113 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R --------------------------------------------------------

Attributel value Ca. example Ca. tag --------------------------------------------------------

Person 1 canto (s.) V/1pi

cantem (pl.) V/4pi --------------------------------------------------------

2 cantes (s.) V/2pi

canteu (pl.) V/5pi --------------------------------------------------------

3 canta (s.) V/3pi

canten (pl.) V/6pi --------------------------------------------------------

Recommendations 114 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R 10.13.5 Gender The expression of Gender is only pertinent in the case of past participle. It is tagged with "m" for masculine and "s" for the feminine.

--------------------------------------------------------

Attribute value Ca. example Ca. tag --------------------------------------------------------

Gender masculine cantat (s.) V/rms

cantats (pl.) V/rmp --------------------------------------------------------

feminine cantada (s.) V/rfs

cantades (pl.)l V/rfp --------------------------------------------------------

subsubsectionNumber The expression of number of past participles has two tags (s and p) corresponding to the two values singular and plural. In the rest of cases, the value of number is entailed by the expression of Number (1-3 for singular, and 4-6 for plural).

--------------------------------------------------------

Attributel value Ca. example Ca. tag --------------------------------------------------------

Number singular cantat V/rms

canto V/1pi cantes V/2pi canta V/3pi --------------------------------------------------------

plural cantats V/rmp

cantem V/4pi canteu V/5pi canten V/6pi --------------------------------------------------------

Recommendations 115 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R 10.14 Application to Irish Among the distinctive features of the Irish verb system are the following: use of verbal nouns and adjectives to achieve distinctions of finiteness, tense, aspect and voice in conjunction with the verb ta' (is); distinction between ta' and the copula is; the presence of a null value on the attribute Person; and a distinction between independent and dependent forms of the verb, the latter used after a variety of verbal particles, including interrogative and negative particles and some complementizers.

10.14.1 Type ===================================================== Attribute values Examples Tags ===================================================== Type Main Scri'obh (se') [(he) wrote] V

Ta' Ta' TA Is Is IS =====================================================

Main verbs are further categorised Transitive or Intransitive. The verb ta' (is) has many of the features of an auxiliary when used in conjunction with verbal nouns and adjectives. Is is a copula requiring that the attribution should be inclusive or total in some sense that is difficult to define. Thus Ta' can be used to say that somebody is tall, fair-haired, etc., but is must be used to say that he/she is Spanish, male/female, a doctor, etc..

10.14.2 Finiteness ===================================================== Attribute values Examples Tags ===================================================== Finiteness finite Scri'obh (se') [he wrote] V

verb.noun scri'obh [writing] VING verb.adj scri'ofa [written] VED =====================================================

This distinction is already recovered, in most cases, in the classification of nouns, verbs and adjectives, and would therefore be redundant in corpus tagging.

10.14.3 Mood ===================================================== Attribute values Examples Tags ===================================================== Mood Indicative Scri'obh (se') [he wrote] VIND

Imperative Scri'odh (se') VINP Condition Scriobhfadh (se') VCON Subjunct (Go) scri'ofa (se') VSUB =====================================================

Recommendations 116 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R 10.14.4 Tense ===================================================== Attribute values Examples Tags ===================================================== Tense Indicative Scri'obh (se') [he wrote] VIND

Imperative Scri'odh (se') VINP Condition Scriobhfadh (se') VCON Subjunct (Go) scri'ofa (se') VSUB =====================================================

As in English, the present tense has habitual aspect. A simple present requires ta' + ag (at) + verbal noun, e.g. Ta' se' ag scri'obh [He is (at) writing]. Further values of Aspect (perfective, imperfect) require composite forms and thus the morphology of Aspect in Irish may be considered complete at this point.

10.14.5 Person ============================================================ Attribute values Examples Tags ============================================================ Person 1 Scri'obh (me')/Scri'obhas VPAST1s

2 Scri'obh (tu')/Scri'obhais VPAST2s 3 Scri'obh (se')/(si') VPAST3sm/f

Scri'obh siad/Scri'obhadar VPAST3p Free Scri'obhadh VPAST0 ============================================================

Person may be marked by an inflection (composite form) or by changing the pronoun in the 3rd singular form, which is always non-composite or split.The so-called free verb is distinctive to Irish. It exists for all tenses and moods. It has some of the features of a passive (Irish verbs do not have the attribute Voice) but can be described in general terms only as a null value of the attribute Person. Thus scri'obhadh leabhar might be glossed a book was written (by somebody) or, equally, somebody wrote a book.

10.14.6 Gender, Aspect, Voice, etc. The other Attributes listed do not apply to Irish.

10.14.7 Language-specific features: The Dependent Form The null person for verbs has been mentioned. The Irish verb also has language-specific attribute Form that may take the values Independent or Dependent. The Dependent form is used with a variety of complementizers and connectives and with particles for interrogation, negation and subjunction. Although this makes the dependent form very common, it generally involves only an initial mutation of the verb, and thus it may be considered part of the syntax of the particles, e.g. go+E, Ni'+L, etc., except in the case of a small number of irregular verbs that have a distinct independent form.

Recommendations 117 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R ============================================================ Attribute values Examples Tags ============================================================ Form independ bhi' [was] TAPAST

dependent Ni' raibh [was not] TAPASTDEP ============================================================

Recommendations 118 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R 11 Adjective

A Type Degr Gen Num Case Use Mod Infl Def Pos Prs Flect M pos m s nom attr pren det

u comp f p gen pred postn indt

l sup n dat advb t acc nomn

voc G qualf comp m s sg 1

e dem sup f p pl 2 n poss ext-c 3

e ord ext-s

l card e indf x inter

excl A indf m s sg 1

l dem f p pl 2 e poss 3

t ord h card N pos m s nom

E comp f p gen R sup mf sp dat

C acc

bas L pos m s nom attr

e comp f p gen pred e sup n dat prem c c acc postm h bas weak

strg

L0 ADJECTIVE

qual pos m sg sg 1 L poss comp f pl pl 2

1 indf sup n 3

card ord L ext-c nom attr prem

2 ext-s gen pred postm a dat

acc L Ge vprt It c It n Gr voc Ge weak Da def Ge flc

2 Ge frct Sp c Sp c Gr ind Ge strg Da indf Ge nof b Ge mxd Da unmk

Recommendations 119 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R 11.1 Comments A large core of agreement emerges as to the Adjective category.

11.1.1 Type The GENELEX and AlethDic models reflect a distinction which is also made in traditional grammars of other Romance languages (e.g. Italian), i.e. a first disjunction between the socalled `qualificatif' and `indicatif' Adjectives. The latter are those Adjectives which also have pronominal function, and in different grammatical traditions are called Determiners in their adjectival function. The former are all the other Adjectives. Indicative Adjectives are further divided into possessives, demonstratives, relatives, indefinites, numerals (ordinals and cardinals), interrogatives, exclamatories. Moreover, the GENELEX model distinguishes 3 possible functions, i.e.:

(i) le chien est n^otre

(ii) n^otre chien

(iii) le n^otre

as (i) Adjective, (ii) Determiner, (iii) Pronoun respectively.

In a preceding version of this proposal we did not propose the feature Type at the recommended level and we had suggested the inclusion of indicative adjectives among Pronouns and/or Determiners, in order to make possible a comparison of the Romance and English traditions. However, after the first cycle of tests and practical applications carried out in the framework of the MULTEXT project, the French partners strongly recommended having the possibility of marking the three possible functions of e.g. possessives, and hence their inclusion also in the Adjective category. Furthermore, the majority of the partners have been in favour of the inclusion in this feature of values such as `cardinal' and `ordinal'. Hence, the values proposed for Type to be marked at common level are: `qualificative', `possessive', `ordinal', 'cardinal' and `indefinite'.

The fact of having indicative adjectives motivates the presence in this table the of features such as Number of Possessor and Person.

Each language-specific application should specify clearly in which category the indicative adjectives are treated in order for cross-linguistic comparisons to be made possible.

11.1.2 Degree GENELEX only foresees, in addition to the values commonly agreed for this attribute, further specifications such as `comp+', `comp-', `comp=' (comparatif-superiorite, comparatif-inferiorite, comparatif-egalite `sup+', `sup-', `subabs', (superlatif-superiorite, superlatif- inferiorite, superaltifabsolu (GENELEX, Sept. 1993), which are proposed here at level 2a, given that they are relevant for many languages.

Recommendations 120 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R 11.1.3 Gender, Number and Case These are arranged in the same way as for Nouns.

11.1.4 Use This is an attribute introduced by MULTILEX among the syntactic specifications, to specify how an adjective can be used: - `attributive': the adjective modifies a noun inside an NP. - `predicative': the adjective can be used as a subject complement of a copular verb, as an object complement or as a secondary predicate. - `adverbial': the adjective modifies a verb or a VP. - `nominalized': the adjective can be used attributively in an NP without head (Dutch-dependent). The value here is not to intended to specify whether a noun can be derived from an adjective.

In the EAGLES rows only the first two are proposed in the common core, Level-2a: they are marked in German and in Dutch.

11.1.5 Modification Type This attribute, called Order in MULTILEX, specifies whether an attributive adjective precedes or follows the noun. The default value can change depending on languages: `prenominal' for English, `postnominal' for Romance languages. The different position often determines distinctions in sense (see section on Italian).

These values are very important in tagging. In the Leech/Wilson proposal, the values of Uses and Modification Type are collapsed in the attribute Uses (in their tagset called Position): the two values `weak' and `strong' are added for German adjective inflection. In the EAGLES proposal these two values are put under the attribute Inflection Type, with the value `mixed' for dealing with the two possibilities. All these values are only relevant for tagging, but are left in the Lexicon proposal as an example of something one might want to record, e.g. in frequency lexicons.

11.1.6 Inflection Type This feature is given for languages such as German.

11.1.7 Definiteness This feature is introduced with the values `definite', `indefinite', `unmark' for Danish.

Recommendations 121 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R 11.2 Application to Italian In Italian, the Adjective agrees in Number and Gender with the noun to which it refers. Another pertinent feature is Degree.

11.2.1 Type Two types are distinguished: `qualificatives' and `determinatives'. The former are tagged `A'; the latter constitute the class of Pronominal Adjectives (see the sections Pronouns and Determiners) and are recognized by the tag `D'. In the Italian Corpus and Dictionary the tag A contains, therefore, the value `qualificative' by default.

11.2.2 Degree The default value is `positive'; adjectives can also have comparative and superlative degree.

- "Analytical forms": The comparatives are, in general, expressed by analytical forms: il piu' bravo, il meno onesto, alto quanto me.

The relative superlative indicates the highest or the least degree of a quality in relation to something (humans or things): il piu' bravo della classe.

These analytical forms present encoding problems from a corpus perspective: they are not dealt with in corpus tagging practice with a word-by-word approach, since they are constructed with more than one word. They belong to the set of phenomena to be codified with the strategy of multi-word expression tagging (see Leech and Wilson Invitation Draft).

-"Synthetic forms". The absolute superlative is constructed by adding the suffixes -issimo, -errimo, -entissimo or the prefixes super-, extra-, iper- etc. to the stem of the adjective: dolcissimo, acerrimo, munificentissimo, ipercritico, ultrarapido.

There are a restricted number of adjectives which have the so-called `organic' comparative and superlative, which are considered as exceptions, e.g. maggiore, migliore, ....

Attribute value It. example It. tag

Degree positive grande (uomo) A/ms

comparative (fratello) maggiore A/msc superlative massimo (dolore)

grandissimo (dolore) A/mss

11.2.3 Gender and Number Within this category, two groups are recognized: - I group: ver-o/-a, ver-i/-e.

Recommendations 122 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R - II group: dolce, dolci. This second group is given the value n (common) for gender in the lexicon, whereas in the corpus it can sometimes be disambiguated by the gender of the noun (see under nouns).

amico vero, amiche vere A/ms A/fp parola dolce, biscotto dolce A/fs A/ms insegnante capace A/ns

Some adjectives are completely invariable: they have the tag nn for Gender and Number in lexicon. Adjectives derived from Adverbs (dappoco, dabbene, etc.) belong to this group.

Attribute value It. example It. tag

Gender masculine vero A/m

feminine vera A/f neuter

l-spec common dolce S/n

Attribute value It. example It. tag

Number singular caro A/ms

plural cari A/mp

l-spec invariant pari A/nn

11.2.4 Use Adjectives can have `attributive' and `predicative' value: these are not encoded in corpus tagging practice, but they are clearly retrievable from the observation of category sequences in a tagged corpus.

However, these two different uses influence the agreement in Number if the adjective refers to more than one noun (Gender is assigned by the Gender of the nouns: if the coordinated nouns are masculine and feminine the adjective selects the value masculine): - predicative use: singular nouns with masculine and feminine gender, mio figlio e mia figlia sono studiosi (Gend: masc.; Numb: plur.). - in attributive use, the agreement in gender is not very precise: with singular nouns which are very similar in meaning, the adjective can agree with the closest noun, i.e. it can be singular: un carattere e una condotta onesta.

11.2.5 Modification Type This information is not marked (but as already pointed out above, it is extractable from sequences).

In prenominal position the adjective looses intensity:

Recommendations 123 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R

un amico caro un caro amico

Always depending on position, some adjectives also change sense:

un uomo povero (i.e. a not rich man) un pover'uomo (i.e. a man to be sympathyzed)

11.2.6 Features not applicable in Italian Case, Inflection Type and Definiteness are not applicable to Italian adjectives.

Recommendations 124 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R 11.3 Application to German 11.3.1 Degree of comparison

Attribute value example tag

Degree positive (das) grosse (Haus) ADJA:Pos.Neut.Nom.Sg.St

(es ist) gross ADJNA:Pos comparative (das) gr"ossere (Haus) ADJA:Comp.Neut.Nom.Sg.St

(es ist) gr"osser ADJNA:Comp superlative (das) gr"osste (Haus) ADJA:Sup.Neut.Nom.Sg.St

(es ist am) gr"ossten ADJNA:Sup

11.3.2 Gender The feature gender applies only to attributive adjectives.

Attribute value example tag

Gender masculine (ein) grosser (Mann) ADJA:Pos.Masc.Nom.Sg.Mix

feminine (eine) grosse (Frau) ADJA:Pos.Fem.Nom.Sg.Mix neuter (ein) grosses (Haus) ADJA:Pos.Neut.Nom.Sg.Mix

11.3.3 Number The feature number applies only to attributive adjectives.

Attribute value example tag

Number singular (der) grosse (Mann) ADJA:Pos.Masc.Nom.Sg.Sw

plural (die) grossen (M"anner) ADJA:Pos.Masc.Nom.Pl.Sw

11.3.4 Case The feature case applies only to attributive adjectives.

Attribute value example tag

Case nominative (der) grosse (Mann) ADJA:Pos.Masc.Nom.Sg.Sw

genitive (des) grossen (Mannes) ADJA:Pos.Masc.Gen.Sg.Sw dative (dem) grossen (Manne) ADJA:Pos.Masc.Dat.Sg.Sw accusative (den) grossen (Mann) ADJA:Pos.Masc.Akk.Sg.Sw

11.3.5 Use The IMS-Tagset considers only the attributive and non-attributive use of adjectives. Adjectives which are used predicatively or as adverbs do not differ in their inflectional shape and are therefore not distinguished. Nominalized adjectives are annotated as common nouns7.

7cf. feature inflection of adjectives (section 11.3.6)

Recommendations 125 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R

Attribute value example tag

Use attributive (ein) schnelles (Auto) ADJA:Pos.Neut.Nom.Sg.Mix

non-attributive (es ist) schnell ADJNA

(es f"ahrt) schnell ADJNA

11.3.6 Inflection The feature inflection applies only to attributive adjectives.

Attribute value example tag

Inflection strong (welch) grosser (Mann) ADJA:Pos.Masc.Nom.Sg.St

weak (der) grosse (Mann) ADJA:Pos.Masc.Nom.Sg.St mixed (ein) grosser (Mann) ADJA:Pos.Masc.Nom.Sg.Mix

Recommendations 126 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R 11.4 Application to English 11.4.1 Degree

Attribute values Examples Tags Degree positive big AJ

comparative bigger AJR superlative biggest AJT

There are no distinctions of Gender, Number or Case among English adjectives. Most adjectives which are disyllabic or polysyllabic, (e.g. dreadful, beautiful) are invariable, and do not take inflection for degree. These adjectives form comparison by the use of the adverbs more and most. There are many exceptions, however, including in particular disyllabic adjectives ending in -y, for which both the inflectional and the non-inflectional comparison are acceptable.

Recommendations 127 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R 11.5 Application to Dutch 11.5.1 Degree of comparison These attribute values also apply to Dutch. In the Lemmas Lexicon the degree of comparison is marked with Yes tags for Positive, Comparative and Superlative. The tags in the table below, however, are single or composed Flection Type tags from the Wordforms tag set. The single tags mark non-inflected forms, the composed tags inflected adjectival forms. Inflection is expressed by the `E' tag, which means: `with suffix `e".

Attribute value Example Tag

Degree positive (een) groot (huis) P

(het) groot (huis) PE comparative (een) groter (huis) C

(het) grotere (huis) CE superlative (dat huis is het) grootst S

(het) grootste (huis) SE

CELEX also attributes the comparative and superlative tags `C' and `S' to some adverbs: `vaker', `vaakst'.

11.5.2 Gender Not applicable in Dutch. (See the contextual Gender tag for Adjectives in nominal use.)

11.5.3 Number Not applicable in Dutch.

11.5.4 Case Not applicable in Dutch.

11.5.5 Use Concerning the attributive-non-attributive use of adjectives, CELEX does not have such tags. But if they existed, they would not be distinctive enough because, according to Dutch grammar, an adjective used predicatively (non-attributively) can still be an adjective or an adverb. In the German non-attributive examples: `er ist schnell' schnell would be an adjective, but in: `es f"ahrt schnell' schnell would be an adverb according to Dutch grammar. CELEX, however, makes a distinction, which is more pertinent to Dutch, between the adverbial and non-adverbial use of adjectives.

The examples in the table below are not taken from the Dutch Linguistic Guide, because they are lacking.

Recommendations 128 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R

Attribute value Example Tag

Use adverbial (hij rijdt) snel adv

non-adverbial (een) snelle (auto) nonadv

This can be an L2 tag. However, a more complete usage table for Dutch Adjectives would be the MULTILEX tag set, presented in the synoptical table of Adjectives. This tagset distinguishes between: Attributive, Predicative, Adverbial and Nominal use:

The table below is a proposal for Dutch, not a CELEX table!

Attribute value Example Tag

Use attributive (de) snelle auto AdjAttrib

predicative (de auto is) snel AdjPred adverbial (de auto rijdt) snel AdjAdv nominal (de/het)snelle AdjNom

11.5.6 Inflection There is no distinction in Dutch between strong, weak and mixed inflection as exists in German. But the Inflection values determined and indetermined , as proposed by MULTILEX, are applicable to Dutch, since the inflection of Dutch adjectival word forms partly depends on whether the article preceding the Adjective is definite or indefinite. A Dutch adjective, used attributively, only has no inflection when preceded by an indefinite article or pronoun and followed by a neuter noun. In all other cases it has inflected forms. Used predicatively, adjectives are always without inflection (with the exception of a certain use of superlatives: `Zijn huis is het mooiste' !). However there is no such Inflection value tag in the CELEX tag set, probably because in the Wordforms Lexicon the P and PE tags of Positive and Positive with suffix `e' cover this domain (see the Degree table).

The table below is a proposal for Dutch, not a CELEX table!

Attribute value Example Tag Inflection determined (het) groote (huis) AdjDet

indetermined (een) groot (huis) AdjIndet

This can be an L2 tag.

Recommendations 129 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R 11.5.7 Language-specific feature: non-verbal part of separable verb We need a special tag to mark adjectives which are part of a separable verb: `fijn' in fijnhakken and `dwars' in dwarsliggen etc.

Recommendations 130 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R 11.6 Application to Spanish 11.6.1 Type Adjectives are not subtyped in ET-ES dictionaries.

11.6.2 Degree

Attribute value Example Tag

Degree positive bueno

comparative mejor superlative buen'isimo

In Spanish most comparatives are expressed by analytical forms; however, we have a few "organic" comparative such as "mejor" or "mayor". Only a small group of adjectives can take superlative suffixes the 'isimo, 'errimo: eg. interesant'isimo celeb'irrimo.

11.6.3 Gender

Attribute value Example Tag

Gender m bueno masc

f buena fem c responsable

There are no neuter adjectives in Spanish. We have adjectives which do not inflect for gender (eg. responsable). Just as in the case of nouns, in the Eurotra dictionaries we leave them unvalued for the attribute Gender. We can, however, add an extra language-specific value (L2b) "common" for these cases.

11.6.4 Number

Attribute value Example Tag

Number s buena sing

p buenas plu i gratis inv

Most Spanish adjectives inflect for number. Only a small group of them are common or invariant.

11.6.5 Modification Type

Attribute value Example Tag

Mod.T. premod un buen libro adjpos=pren

postmod un libro bueno adjpos=postn indif largo adjpos=none

Recommendations 131 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R Some Spanish adjectives are restricted wrt order (in some cases order determines the meaning of adjectives: un pobre hombre vs. un hombre pobre). We have adjectives which:

a) can only occur as postmodifiers: "una entrada accesible" b) can only occur as premodifiers: "un buen/gran hombre" (most of these adjectives have their postmodifier form: "un hombre bueno/grande", (apocope)). c) can occur as both pre- and postmodifiers: "un (dif'icil) trabajo (dif'icil).

These facts are coded in ET-ES grammars under the feature "adjpos".

Recommendations 132 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R 11.7 Application to French (Corpus) 11.7.1 Type As indicated before, French has many more adjective types than just qualificative. These will be dealt with in the pronoun-determiner section.

11.7.2 Gender

Attribute value Example Tag Gender masculine petit ADJEMS

feminine petite ADJEFS

11.7.3 Number

Attribute value Example Tag Number singular grand ADJEMS

plural grands ADJEMP

11.7.4 EAGLES features not applicable Use: As mentioned for other languages, attributive/predicative use is a syntactic distinction which applies to French. Since all adjectives can be used either in attributive or in predicative position and would not allow for other disambiguations, the distinction is not relevant in a tagset. Degree applies to French, but is marked by external premodifiers and therefore does not require a special adjective class in the tagset. Modification type applies to French; in some specific cases, the position even implies a semantic distinction: e.g. un grand homme (a great man) vs. un homme grand (a tall man). Although the distinction could be of interest in a tagset for its prediction potential, it is not used in the IBMF tagset. Case and inflection type do not apply to French. Definiteness does not apply, although it applies for other types of French `adjectives" (see determiners).

11.7.5 IBMF Tagset features not applicable in EAGLES The tagset has a special feature for indefinite adjectives. This will be put in the pronoundeterminer class.

11.8 Application to French (Lexicon) 11.8.1 Type

============ =========== =========== ==== Attribute Value Example Code ============ =========== =========== ==== Type qualificat. bon f

ordinal deuxi'eme o cardinal deux c

Recommendations 133 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R

indefinite quelconque i possessive mien s ------------ ----------- ----------- ----

11.8.2 Degree ------------ ----------- ----------- ---- Attribute Value Example Code ------------ ----------- ----------- ---- Degree positive bon p

comparative meilleur c ------------ ----------- ----------- ----

Note: The distinction positive/comparative applies only to two adjectives in French: bon and mauvais. All other adjectives form their comparatives with plus + adjective (e.g., plus grand). Superlative is also a compound form (le + comparative, e.g. le plus grand).

Recommendations 134 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R 11.8.3 Gender ------------ ----------- ----------- ---- Attribute Value Example Code ------------ ----------- ----------- ---- Gender masculine bon m

feminine bonne f ------------ ----------- ----------- ----

11.8.4 Number ------------ ----------- ----------- ---- Attribute Value Example Code ------------ ----------- ----------- ---- Number singular bon s

plural bons p ------------ ----------- ----------- ----

11.8.5 Case Not applicable to French.

11.8.6 Combinations --------- ----------- Tag Example --------- -----------

========= ======= ============================================= Lexique Corpus Example ========= ======= ============================================= Afcfp- AFP meilleures Afcfs- AFS meilleure Afcmp- AMP meilleurs Afcms- AMS meilleur Afpfp- AFP bonnes Afpfs- AFS bonne Afpmp- AMP bons Afpms- AMS bon

Ai-fp- AFP certaines, memes, quelconques Ai-fs- AFS certane, meme, quelconque Ai-mp- AMP certain, memes, quelconques Ai-ms- AMS certain, meme, quelconque

Ac-fp- AFP deux Ac-fs- AFS une Ac-mp- AMP deux

Recommendations 135 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R Ac-ms- AMS un Ao-fp- AFP premieres Ao-fs- AFS premiere Ao-mp- AMP premiers Ao-ms- AMS premier

As-fp- AFP leurs, miennes, tiennes, siennes, notres, votres As-fs- AFS leur, mienne, tienne, sienne, notre, votre As-mp- AMP leurs, miens, tiens, siens, notres, votres As-ms- AMS leur, mien, tien, sien, notre, votre =====================================================================

Recommendations 136 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R 11.9 Application to Portuguese 11.9.1 Type ------------ --------------- ----------- ---- Attribute Value Example Tag ------------ --------------- ----------- ---- Type qualificative azul

possessive seu indefinite cardinal dois ordinal primeiro ------------ --------------- ----------- ----

11.9.2 Degree ------------ --------------- --------------- ---- Attribute Value Example Tag ------------ --------------- --------------- ---- degree positive

comparative melhor superlative paupe'rrimo ------------ --------------- --------------- ----

In the Portuguese model the `positive' value is not explicitly marked because it is considered a default value.

11.9.3 Gender ------------ ----------- ----------- ---- Attribute Value Example Tag ------------ ----------- ----------- ---- Gender masculine bom

feminine boa neuter ------------ ----------- ----------- ----

The value neuter does not apply to Portuguese adjectives.

11.9.4 Number ------------ ----------- ----------- ---- Attribute Value Example Tag ------------ ----------- ----------- ---- Number singular bom

plural bons ------------ ----------- ----------- ----

Recommendations 137 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R 11.9.5 Case Not applicable to Portuguese.

11.9.6 Other features required to encode Possessive Adjectives The following features are also required to encode the Portuguese possessive adjectives:

11.9.7 Possessor ------------------------------------------------------ Attribute Value Example Tag ------------------------------------------------------ Number-possessor singular meu

plural nosso ------------------------------------------------------

11.9.8 Person As we explained in the section on Verbs, the Portuguese lexicon adopted the decomposition of the feature person into the features person-deixis and person-conc. These features are also used to encode Portuguese possessive adjectives.

Recommendations 138 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R 11.10 Application to Danish For Danish adjectives the following features are relevant: Type, Degree, Gender, Number, Use and Definiteness (language-specific property). The feature Case applies only to nominalised adjectives (cf. section Noun, item Case.)

11.10.1 Type The attribute Type, including cardinals, ordinals and quantifiers, seems to be difficult to treat cross-linguistically. Basically, in Danish these behave in a similar way to adjectives, but within the Eurotra framework they were treated as separate classes (although they were previously regarded as sub-categories of the adjective category.)

11.10.2 Degree The default value of the attribute Degree is `positive' (called `base' in EDEMD). The absolute superlative can be regarded as an additional degree; it is composed of the prefix aller- and the superlative of the adjective. There are a huge number of adjectives which have periphrastic instead of inflectional comparison; in such cases the comparative is comprised of mere + the positive form of the adjective, and the superlative of mest + the positive form. In corpus annotation the analytical forms occurring in the text are recognised word-by-word, i.e. not recognised as comparative or superlative.

The EDEMD applies the attribute `Comparison type' to all adjectives; this feature is relevant to lexicons only. Traditional dictionaries of Danish indicate deviations from the regular inflectional patterns.

Attribute value Example Tag

Degree positive dyr adj pos

comparative dyrere adj comp superlative dyrest adj sup

11.10.3 Gender, Number and Definiteness Agreement applies to adjectives in gender, number and definiteness, when the adjective is used in attributive function. Definiteness does not apply in predicative function. A number of adjectives have defective inflection. Another group have no possibility of inflection (gender, number and definiteness) or comparison: they are completely invariable.

Gender

Attribute value Example Tag

Gender common (en) dyr (bil) adj com

neuter (et) dyrt (hus) adj neut

Recommendations 139 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R Number The adjective has no gender and definiteness agreement in the plural (neither in attributive nor in predicative functions). All regular adjectives receive -e as the plural ending, both in definite and in indefinite use.

Attribute value Example Tag

Number singular (en) dyr (bil) adj com sg

plural dyre (biler) adj pl

Definiteness The morphological distinction between definite and indefinite only exists in the singular in attributive function. As mentioned above (item Number), the definite and indefinite plural forms of adjectives are identical. However, a parser-based corpus tagger can distinguish between the two forms on the basis of the immediately preceding context of the adjective.

Attribute value Example Tag Definiteness indefinite (en) dyr (bil) adj com sg indef

definite (den) dyre (bil) adj com sg def Definiteness indefinite dyre (biler) adj pl indef

definite (de) dyre (biler) adj pl def

11.10.4 Use Corpus tagging devices recognise the morphological forms; however, in the case of syncretic forms (e.g. singular definite and plural) it is necessary to parse the sentence to disambiguate the form or else the tagger may insert more than one single tag.

Attribute value Example Tag

Use attributive (en) dyr bil adj ... attr

predicative (bilen er) dyr adj ... pred adverbial (pigen taler) smukt adj ... adv nominal (den/det/de) dyre adj ... nom

11.10.5 Synoptic table of agreement relations The value of Use may also be inserted into the tag combination if needed for special purposes.

Recommendations 140 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R

Attribute value Example Tag Agreement A Gender, Number, Def.ness en dyr bil adj com sg indef

et dyrt hus adj neut sg indef den dyre bil adj com sg def det dyre hus adj neut sg def Number, Def.ness dyre biler/huse adj pl indef Number, Def.ness (de) dyre biler/huse adj pl def

Agreement P Gender, Number en bil/bilen er dyr adj com sg

et hus/huset er dyrt adj neut sg Number biler/husene er dyre adj pl

A word-by-word tagger is able to annotate the plural adjective form with the definiteness feature on the basis of the immediate context (i.e. presence or absence of articles, possesives, etc.)

Recommendations 141 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R 11.11 Application to Greek Features that apply to Greek adjectives are: Degree, Gender, Number and Case.

11.11.1 Degree

Attribute value Gr. example Gr. tag

Degree basic psylos AjBa

comparative psyloteros AjCp superlative psylotatos AjSu Almost all adjectives have a form for the value basic. Some of them form the comparative and the superlative degree in one of the following ways:

(a) periphrastically, on the basis of the basic form and a word/phrase denoting comparison:

Comparative: pio kalos, lighotero kalos, to idhio kalos

Superlative: o pio kalos, o lighotero kalos

This type of formation can only be dealt in corpora by multi-word tagging, while no information on this is to be coded in the lexicon.

(b) Certain adjectives can also form the comparative degree (denoting superiority) by the addition of -ter- between the stem and the ending, and the superlative by -tat-:

Basic: puknos, psylos Comparative: puknoteros, psyloteros

Superlative: puknotatos, psylotatos

(c) A few adjectives form the comparative and superlative degree monolectically, using a different stem or a different ending from the normal one:

Basic: kalos, meghalos, aplos Comparative: kaluteros, meghaluteros, aplousteros

Superlative: kallistos, meghistos, aploustatos

A few adjectives are found only in the comparative and superlative degree; this is explained by the fact that they are derived from the basic form of an adverb:

Basic: anw, katw - adverb Comparative: anwteros, katwteros - adjective

Superlative: anwtatos, katwtatos - adjective

Recommendations 142 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R 11.11.2 Gender, Number and Case In Greek, adjectives must agree in gender, number and case with the nouns they modify.

The majority of Greek adjectives is inflected according to paradigms that form distinct forms for all three genders. A few adjectives (e.g. ending in -ys, -ys -es) share the same form for the masculine and the feminine gender, and another category (mainly foreign words) share the same form for all genders:

kalos, kaly, kalo

but eutuhys, eutuhys, eutuhes

and roz, roz, roz

Although for nouns sharing the same form for the masculine and feminine gender we use the value masc-fem, we have decided not to use it for adjectives for two reasons: a) not all inflected forms of the masculine and feminine gender coincide, b) in the majority of cases, the value can be disambiguated during corpus annotation on the basis of the noun they refer to:

O Ghiannys einai eutuhys - masc.

Y Maria einai eutuhys - fem.

On the other hand, we introduce the value common for those adjectives that have a common form in all genders. No disambiguation is attempted at this point during corpus processing, as it would complicate the procedure too much. In relation to Number and Case, the values invariant and indeclinable respectively are used for adjectives that retain the same form irrespective of gender, number and case. As in the case of nouns, these are foreign words which have entered the Greek language without having adopted the inflectional system, and their disambiguation relies on context.

Attribute value Gr. example Gr. tag

Gender masculine kalos AjBaMa

feminine kaly AjBaFe neuter kalo AjBaNe l-spec common roz AjBaCo

Attribute value Gr. example Gr. tag

Number singular kalos AjBaMaSg

plural kaloi AjBaMaPl l-spec invariant roz AjBaMaSgNv

Recommendations 143 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R

Attribute value Gr. example Gr. tag

Case nom kalos AjBaMaSgNm

gen kalou AjBaMaSgGe acc kalo AjBaMaSgAc dat kalw AjBaMaSgDt l-spec voc kale AjBaMaSgVo l-spec indcl roz AjBaMaNvIc

11.11.3 Use Although this feature (belonging to level 2a) is not currently used in corpus tagging, it is applicable to Greek with the values proposed by EAGLES, i.e. attributive and predicative.

11.11.4 Comments Type is not used for Greek, given that cardinal and ordinal adjectives are included under the category of num. Details on the reasons that dictate this practice are given in the relevant section.

11.11.5 Features not applicable in Greek Modification Type, Inflection Type and Definiteness are not applicable to Greek adjectives.

Recommendations 144 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R 11.12 Application to Swedish For Swedish adjectives the following features are involved; Type, Degree, Gender, Number, Use, Case and Definiteness.

Type Use Degree Definiteness Number Gender Case

(masc.) normal positive singular

attribute indefinite nonneuter attributive comparative

predicate definite neuter predicative superlative plural ordinary

nominative (nonmasc) genitive

11.12.1 Type Type is here, as elsewhere, used for "syntactic type". This is a lexically determined feature, governing the syntactic use of the adjective. The values are normal, attributive and predicative. (Other proposed classifications subsumed under Type are superfluous for Swedish and seen to be semantically, rather than morphosyntactically motivated.)

Attribute value Example Tag Type normal trevlig adj

attributive veritabel adj-attr predicative slkt adj-pred

11.12.2 Use This feature takes the values attribute, predicate and nominalization (note the differences to the values of Type). Adjectives of Type normal can take any of these values, those of Type attribute can take values attribute and nominalization, those of Type predicative only predicate.

Attribute value Example Tag Use attribute (en) dyr bil adj-...-attr

predicate (bilen "ar) dyr adj-...pred nominalization (de) dyra adj-...-nom

11.12.3 Degree This feature is not applicable to all adjectives, nor can it always be morphologically realized, even when applicable. The values are positive, comparative and superlative. Distinctions in Definiteness, Number and Gender can not be realized for comparative and superlative (the latter is, of course, inherently definite).

Recommendations 145 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R

Attribute value Example Tag Degree positive fin adj-pos

comperative finare adj-comp superlative finast adj-sup

11.12.4 Definiteness In attribute Use, the value is governed by the head of the noun phrase in question. The values are indefinite and definite, where indefinite is prescribed for predicate Use and definite for nominalization Use. NB that when definite is realized, Number and (normal) Gender distinctions are neutralized.

Attribute value Example Tag Definiteness indefinite (en) fin (stol) adj-sg-indef

definite (den) fina (stolen) adj-sg-def Definiteness indefinite (flera) fina (stolar) adj-pl-indef

definite (de) fina (stolarna) adj-pl-def

11.12.5 Number The values are singular and plural. If plural is realized, Gender distinctions are neutralized.

Attribute value Example Tag Number singular (en) fin (stol) adj-com-sg

plural (flera) fina (stolar) adj-pl

11.12.6 Gender There are two pairs of values for this feature, i.e. on the one hand nonneuter vs neuter, on the other hand masculine vs nonmasculine. The first pair is only applicable to indefinite adjectives and is governed by the grammatical gender of the head of the noun phrase or the subject of the clause. The second pair is only applicable to definite adjectives (not superlatives) and is governed by the semantic gender of the head of the noun phrase, or by the natural sex of the underlying concept, in the case of nominalizations.

Attribute value Example Tag Gender nonneuter (en) fin (stol) adj-non

neuter (ett) fint (bord) adj-neut masculine (den) gamle (mannen) adj-masc nonmasculine (den) gamla (kvinnan) adj-nmasc

11.12.7 Case This section is applicable only for adjectives in nominalizations. The values are those of nouns, i.e. ordinary and genitive.

Recommendations 146 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R

Attribute value Example Tag Case ordinary (de) unga adj-ord

genitive (de) ungas adj-gen

Recommendations 147 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R 11.13 Application to Catalan 11.13.1 Type Adjectives are not specifically subtyped. The only specific tag distinguishes the cardinal adjectives from the rest.

--------------------------------------------------------

Attribute value Ca. example Ca. tag --------------------------------------------------------

Type general bonic A/ms --------------------------------------------------------

cardinal tres AN --------------------------------------------------------

11.13.2 Degree Positive and comparative are not tagged. In Catalan there is just a very small group of adjectives that can have morphologically-based comparatives. The usual form of comparisons is syntactically based. Superlatives take the form tag "d" in the first position of its inflectional specifications, and so do the other adjectives formed with non-lexical (aspectual) suffixes.

--------------------------------------------------------

Attribute value Ca. example Ca. tag --------------------------------------------------------

Degree sup bonissim A/dms --------------------------------------------------------

bonet A/dms --------------------------------------------------------

11.13.3 Gender The gender of adjectives is marked with the form tags "m" of "f". Invariable adjectives don't take any gender tag. In Catalan there is no neuter adjectives. A gender value "c" 'common' is needed for the gender (and sometimes gender and number) invariable adjectives, in the same way of L2b value proposed for the Italian and Spanish.

--------------------------------------------------------

Attribute value Ca. example Ca. tag --------------------------------------------------------

Gender m bo A/ms --------------------------------------------------------

f bona A/fs --------------------------------------------------------

c possible AI/s --------------------------------------------------------

c gaire AII --------------------------------------------------------

Recommendations 148 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R 11.13.4 Number Most Catalan adjectives inflect in number, only a subgroup of gender invariable adjectives are also number invariable.

--------------------------------------------------------

Attribute value Ca. example Ca. tag --------------------------------------------------------

Number s bona A/fs --------------------------------------------------------

p bones A/fp --------------------------------------------------------

p possibles AI/p --------------------------------------------------------

Case, Inflection Type and Definiteness are not pertinent. We don't have specific marks for Use and Modification Type.

Recommendations 149 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R 11.14 Application to Irish The so called indicative adjectives (possessive, demonstrative, and ordinal adjectives) are treated as determiners on syntactic grounds, and likewise indicators of quantity and reference-determiners based on discourse properties, such as ce'anna (same), eile (other). The broader treatment of adjectives, which would includes the indicators as a special type, is possible also, and is in fact the standard presetation in the grammar books. By taking the restrictive view of adjectives, as is done in Multilex, the only attributes that need be considered are Degree, Gender, Case and Number.

11.14.1 Degree ============================================================ Attribute values Examples Tags ============================================================ Degree Base fuar (cold) ADJ

Comparat ni'os fuaire (colder) ADJCOM Superlat is raibh [was not] TAPASTDEP ============================================================

11.14.2 Gender ============================================================ Attribute values Examples Tags ============================================================ Gender Masculine (la') fuar (a cold day) ADJm

Comparat (oi'che) fhuar (a cold night ADJf ============================================================

11.14.3 Number ============================================================ Attribute values Examples Tags ============================================================ Number Singular (la') fuar (a cold day) ADJms

Plural laethanta fuara (cold days) ADJmp ============================================================

11.14.4 Number ============================================================ Attribute values Examples Tags ============================================================ Case Nominative (la') fuar (a cold day) ADJms

Genitive laethanta fuara (cold days) ADJmp Vocative laethanta fuara (cold days) ADJmp ============================================================

Recommendations 150 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R 11.14.5 Other attributes Attributive, predicative, adverbial and nominal uses of adjectives exist. There is no morphological change, however, and the Use category can always be recovered from the syntax. Modification type is postnominal. Inflection type and indefiniteness do not apply.

Recommendations 151 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R 12 Pronoun

P Type wh-T Pers Gend Numb Case Pos Pol Funct Infl M dem 1 m s nom

U indf 2 f p gen

L pers 3 n dat T poss acc

rel voc G pers 1 m s sp weak

E rel 2 f p pl strg N poss 3 n

E indf

L excl

imprs A dem 1 m s sg

l poss 2 f p pl e rel 3 n

t pers h indf

imprs N poss int 1 m s nom

E dem rel 2 f p gen R indf 3 mf sp dat

C int/rel acc

pers obl refl bas L poss int 1 m s nom

e dem rel 2 f p gen e indf 3 n dat c int/rel c acc h pers obl

refl bas

E-L0 PRONOUN

E dem 1 m sg nom sg A indf 2 f pl gen pl G poss 3 n dat

int acc rel obl L pers obj

1 refl

recp excl L int/rel pol

2 pres/ref fam a dem/indf L It c It n Po pobj Du att Fr/It weak

2 Du prd Fr/It strg b Du adv Ge infl

Ge no-infl

Recommendations 152 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R 12.1 Comments The NERC and the Leech/Wilson schemes propose a multilayered treatment of Pronouns, thus permitting different levels of granularity in annotation. The reader should note that the table above describes the most fine-grained level of linguistic description of the two proposals, where Pronouns are recognized as a category of their own.

At a less granular level, in these two systems Pronouns appear merged together with Determiners. The background reason for this merging in the NERC survey was the necessity of meeting the requirements, on the one hand, of a number of English tagsets (e.g. Penn Treebank, Brown, Lancaster) where, for example, Demonstratives are undistinguished as to their pronominal and determiner functions and receive a unique tag, and, on the other hand, of tagsets which include Articles among the Determiners. Hence, a multilayered approach was adopted, where three different fine-grained levels of linguistic distinctions offer the possibility for each existing practice to be placed at the appropriate level, thus permitting its reusability (see Monachini and Oestling 1992b). This solution has also been adopted by the EAGLES Corpus group for linguistic annotation (Leech and Wilson 1994).

A first version of the present document also proposed the same treatment of Pronouns and Determiners. However, after the first cycle of applications, and in particular after the MULTEXT concrete testing, it seemed better to distinguish between different functions and, therefore, to have different categories for Pronouns and Determiners, at least at the lexical level. Lexical descriptions should be independent from applications and should aim at a general description of each language; corpus tags, depending on the capabilities of state-of-art tagging techniques, may underspecify lexical specifications, collapsing many distinctions and presenting broader categories (Calzolari and Monachini 1994). Furthermore, following the TEI proposal, it has also been decided to have Articles as a separate category.

12.1.1 Type, Wh-Type The column Type shows how Pronouns are split into different subclassifications, recommended at Level-1. The subfeature wh-Type is for the further distinction of the double value `interrogative/relative' and is proposed at Level-2a.

12.1.2 Person, Case and Possessor Other information, such as Person, Case and Possessor (i.e. the Number of Possessor), are clearly not applicable to all the Types. This tabular representation does not permit the indication of constraints on the application of some features in the presence of others. These constraints have to be explicitly specified in the language-specific applications.

Recommendations 153 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R 12.1.3 Politeness This encodes the polite usage of personal pronouns.

12.1.4 Function This feature has been added for the encoding of `attributive', `predicative' and `adverbial' use (see Dutch specific application.)

12.1.5 Other values on Level 2b Finally, on level 2b, the value `prepositional obj' is foreseen among the values of the attribute Case, for the Portuguese language.

Recommendations 154 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R 12.2 Application to Italian In Italian, both corpus and dictionary distinguish Pronouns according to the feature Type (e.g. indefinite, demonstrative, possessive, etc.).

12.2.1 Type The table below shows the different types pertinent to the category of Pronouns. In the following sections, each Type will be discussed in detail.

Attribute value It. example It. tag

Type dem questo PD/ms

poss mio PP/ms indf ognuno PI/ms pers io PP/ns1s refl si PF/nn wh-Type int che PT/ns

rel che PR/ns excl quanto! PE/ns

12.2.2 Personal Pronoun Personal Pronouns are inflected for Person and Number, as shown in the following table.

Personal example It.tag Person Gend.-Numb. 1 s io PP/ns1s 2 s tu PP/ns2s 3 s egli PP/ms3s 1 p noi PP/np1s 2 p voi PP/np2s 3 p essi PP/mp3s

As far as the pronominal paradigm is concerned, Case is not encoded at present in our DMI and corpus. Personal pronouns are not lemmatized: `gli' is not considered the dative form of the base pronoun `egli' (he), but constitutes a separate entry.

The Italian pronominal paradigm is described below: forme toniche: subj (`io, egli'), compl (`me, lui') ama me / da' a me (dir-obj/prep-obj) (he loves me / he gives to me)

Recommendations 155 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R ama lui / da' a lui (dir-obj/prep-obj) (she loves him / she gives to him)

forme atone: --, compl (`mi, gli/lo') mi da' / mi ama (ind-obj/dir-obj) (he gives me / he loves me) gli da' (ind-obj) (he gives him) lo ama (dir-obj) (she loves him)

This paradigm can be mapped on the proposed Case system in the following way:

io, egli subj nom mi/me dir-obj/ind-obj/prep-obj obj = acc, dat, prep+obj lui dir-obj/prep-obj obj = acc, prep+obj gli ind-obj dat lo dir-obj acc

Polite form In Italian, the modern system of Personal Pronoun for addressing a person is bipartite: tu and lei, feminine singulal used for the polite form (also for addressing masculine persons). Polite usages are very interesting from a corpus perspective, but, at present, are not encoded in our tagger.

In polite usages, two different types of agreement can be used: - agreement by nature: the participle has masculine gender, i.e. agrees with the masculine noun and not with the personal pronoun. Professore, Lei si e' occupatO. Professor (man), She has been interested in ... Professor, you have been interested in ...

- agreement by grammar: the participle is feminine, i.e. agrees with the pronoun, even though referred to a masculine noun. Lei, Professore, l'ho sempre ascoltatA Professor (man), I always have listened to Her Professor, I always have listened to you

Non-tonic personal pronouns always agree by grammar: Professore, vorrei dirLE, spero di rivederLA presto Professor (man), I want to tell Her, I hope to see Her soon. Professor, I want to tell you ... , I hope to see you soon.

Recommendations 156 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R 12.2.3 Reflexive (Pronoun) In Italian, the reflexive pronouns are represented by mi, ti, si, se', ci, vi, si. They are inflected for Person and Number, while Gender is not pertinent.

For the third singular and plural persons, in addition to si, the tonic form se' can be used: egli si lava; egli lava se' (stesso); esse aiutano solo se' (stesse). Se' can be `reinforced' by the ajdective stesso: in this case the bigram se stesso should be encoded on the basis of multiword expression tagging strategy.

Reflexive Gender Number example It.tag 1 s mi PF/nn1s 2 s ti PF/nn2s 3 s si, se' PF/nn3s 1 p ci PF/np1p 2 p vi PF/np2p 3 s si PF/nn3p

Formally, reciprocal pronouns are the same as reflexive pronouns. 12.2.4 Possessive Possessives are inflected for Number and Gender and agree with the nouns to which they refer; they are distinguished according to the Person to which they refer.

Possessive

Person 1s 2s 3s 1p 2p 3p Gender Number m s mio tuo suo nostro vostro loro m p miei tuoi suoi nostri vostri loro f s mia tua sua nostra vostra loro f p mie tue sue nostre vostre loro

Scrivo con la tua penna, perche' non ho la mia PP/fs Italian has two more possessives: altrui (of other people) and proprio (own), which can both function as pronouns:

spende il denaro altrui, non il proprio PP/nn occorre dare del proprio denaro, non dell'altrui PP/nn

Recommendations 157 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R 12.2.5 Possessor Information about the possessor is not encoded in the Italian lexicon or corpus, but it can be inferred from the lemma:

queste sono le mie (owned things plural, but possessor singular), is tagged PP/fp questa e' la nostra (owned thing singular, but possessor plural) is tagged PP/fs

12.2.6 Demonstrative

Demonstrative example It.tag

questo PD/ms quello PD/ms

Among the Demonstratives, the most used are: questo, codesto, quello. Deixis is implicitly contained in demonstratives, but not presently encoded:

usa queste PD/fs, non prendere quelle PD/fp While the others can be used as pronouns or pronominal adjectives, cio' only has pronoun function:

cio' e' vero PD/nn Questi, quegli, costui, colui are pronouns used only for humans; questi and quegli, scarcely used and archaic, refer to the singular:

Questi e' un bell'uomo PD/ms Stesso, medesimo can be demonstrative pronouns.

12.2.7 Indefinite

Indefinite example It.tag

ognuno PI/ms chiunque PI/ns

Italian Indefinites are inflected for Gender and Number. Among the indefinites, the following can only be pronouns: uno, ognuno, qualcuno, chiunque, chicchessia, nulla, niente.

ognuno deve fare ... PI/ns Some can have both the pronoun and the pronominal adjective/determiner functions: alcuno, ciascuno, taluno, nessuno, tutto, alquanto, poco, molto, troppo, tanto.

ho comprato il libro e l'ho letto tutto PI/ms

Recommendations 158 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R 12.2.8 Interrogative

Interrogative example It.tag

chi PT/ns quale PT/ns

Interrogatives are inflected for Gender and Number. Chi is pronoun only.

Chi viene oggi? PT/ns Che, quale, quanto can be either pronouns or pronominal adjectives/determiners:

A che pensi? PT/nn

12.2.9 Exclamatory

Exclamatory example It.tag

quanto! PE/ms quale! PE/ns

Che, quale, quanto can also have exclamatory value:

quanti sono venuti! PE/mp

12.2.10 Relatives

Relative example It.tag

che PR/nn

The Relatives comprise che, il quale, cui. Che is used both for singular and plural, and for masculine and feminine.

Recommendations 159 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R 12.3 Application to German 12.3.1 Type

Attribute value example tag

Type personal ich PPER:1.Sg.Nom

reflexive sich PRF:3.Sg.Akk possessive meins PPOSS:Neut.Nom.Sg demonstrative dieses PDEMS:Neut.Nom.Sg relative , das PRELS:Neut.Nom.Sg indefinite irgendeines PROS:Neut.Nom.Sg interrogative was? PWS:Neut.Nom.Sg

12.3.2 Gender The feature gender applies to all pronouns - except personal pronouns (where it is relevant only for 3rd person singular) and reflexive pronouns. It is not defined for a number of indefinite pronouns such as jemand, niemand, man etc.

Attribute value example tag

Gender masculine er PPER:3.Sg.Nom.Masc

feminine meine PPOSS:Fem.Nom.Sg neuter das PRELS:Neut.Nom.Sg

12.3.3 Number The feature number applies to all pronouns.

Attribute value example tag

Number singular keiner PROS:Masc.Nom.Sg

plural welche PWS:Neut.Nom.Pl

12.3.4 Case The feature case applies to all pronouns.

Attribute value example tag

Case nominative keiner PROS:Masc.Nom.Sg

genitive niemandes PROS:Gen.Sg dative jemandem PROS:Dat.Sg accusative wen? PWS:Akk.Sg

12.3.5 Possessor The feature possessor is not encoded in the IMS-Tagset. It can be determined from the lemma.

Recommendations 160 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R 12.4 Application to English 12.4.1 Pronoun-type

Attribute values Examples Tags Pronoun-type personal it, she PPs3, PPs3NF

reflexive myself PRs1 possessive yours PV2 demonstrative this PDs wh-type what PW indefinite anyone PI

12.4.2 Wh-subtype

Attribute values Examples Tags Wh-subtype relative which PWR

other which PWQ

The relative and "other" wh-type pronouns may be distinguished by the following examples: the shoes which she bought (PWR); Which shoes did she buy? (PWQ).

The "other" (non-relative) category includes interrogative and exclamatory pronouns and determiners. As these subtypes are difficult to distinguish automatically, it is convenient to bring them together under a single value "other".

12.4.3 Person

Attribute values Examples Tags Person 1st person I PPs1N

2nd person you PP2 3rd person she PPs3NF

12.4.4 Number

Attribute values Examples Tags Number singular someone, it PIs, PPs3

plural few, they PIp, PPp3N

12.4.5 Gender

Attribute values Examples Tags Gender masculine him PPs3OM

feminine her PPs3OF neuter its PVs3U common anyone PIs

The common gender pronouns such as anyone have personal reference, but are neutral between masculine and feminine.

Recommendations 161 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R 12.4.6 Case

Attribute values Examples Tags Case nominative they PPp3N

oblique them PPp3O

Personal pronouns in the oblique case (me, them, etc) are used as objects, as prepositional complements, and also in some other functions.

Recommendations 162 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R 12.5 Application to Dutch 12.5.1 Type CELEX distinguishes 10 subclasses of pronouns:

Attribute value Example Tag

Type personal ik PRON pers

reciprocal elkaar PRON wkg. reflexive zich PRON wknd. possessive mijn PRON bez. demonstrative dit PRON aanw. relative wat PRON betr. indefinite geen PRON onbep. interrogative wie? PRON vraag. exclamatory wat! PRON uitr.

12.5.2 Gender A gender value tag set is applicable to Dutch for Personal, Demonstrative, Possessive, Interrogative, Reflexive and Relative Pronouns, but is not found in the CELEX tagset. The table below is a proposal for Dutch, not a CELEX table !

Attribute value Example Tag

Gender masculine zijn (vader) Masc

feminine haar (vader) Fem neuter welk (boek) Neut

Gender of pronouns is very important for contextual disambiguation of gender ambiguities. See Noun.

12.5.3 Number Number is applicable to Dutch Articles and Demonstrative, Reflexive, Relative and Interrogative Pronouns, but is not in the CELEX tagset.

The table below is a proposal for Dutch, not a CELEX table !

Attribute value Example Tag

Number singular het (huis) Sing

(het huis) dat Sing plural de (huizen) Plur

(de huizen) die Plur

Recommendations 163 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R 12.5.4 Case Case is only applicable to Dutch in some archaic forms still surviving in the language in idiomatic expressions (Articles, Relative, Personal and Interrogative Pronouns). It is only present in the CELEX Wordforms Lexicon. The tags are the same as for Nouns:

Ge: Genitive singular: des, dezer, harer etc. Gm: Genitive plural: aller, hunner, onzer De: Dative singular: aller, den, der, dien Dm: Dative plural: haren, hunnen, mijnen, onzen, uwen, zijnen.

12.5.5 Function There is no Function value tag set in CELEX. The table of the German application is not applicable to Dutch.

The table below is a proposal for Dutch, not a CELEX table ! The feature Function does not apply to all pronouns.

Attribute value Example Tag Function attributive deze (auto) Attrib

predicative (Dat is) het Pred adverbial Hoe (werkt dat?) Adv nominal (Het) mijine Nom

The value Predicative might prove redundant since predicatively used pronouns might turn out to be only nominal. This can be an L2 tag.

Recommendations 164 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R 12.6 Application to Spanish The Eurotra dictionaries count pronouns as nouns. Among all the attributes used to describe nouns the following refer to pronouns: - "Nform", which serves to distinguish between pronouns (valued pro or cli) and "normal" nouns (valued norm).

- "dtype", which helps to distinguish between possesive and non-possesive pronouns. - "whmor", used to distinguish between relative pronouns (valued as rel) and interrogative pronouns (valued as int) from the rest of the nominals (valued none).

By means of these three attributes we can distinguish between:

Ex. Nform Dtype Whmor "class" casa normal non-poss none "normal" noun

yo pro non-poss none personal pro este pro non-poss none demonstrative pro alg'un pro non-poss none indefinite pro

mi pro poss none possesive pro mio pro poss none possesive pro cuyo pro poss rel relative pro quien pro non-poss rel relative pro qui'en pro non-poss int interrogative pro

me cli non-poss none clitic

12.6.1 Type Demonstrative: PD

Only the attributes Number and Gender are pertinent to Spanish pronominal demonstratives.

Pers Number Gender Pos Case Funct Pol Infl Ex.

3 sg masc este 3 pl fem aquellas 3 sg n eso

These pronominals reflect deictic degree of remoteness, which is not, however, coded in ET-ES dictionaries.

As for "person", these pronominals (coded in ET-ES dictionaries as "nouns") are forced to have a person attribute valued as "third", due to verbal agreement.

Possessive: PP

Recommendations 165 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R The attributes pertinent to Spanish pronominal possesives are: Number, Gender, Person and Possessor.

Pers Number Gender Pos Case Funct Pol Infl Ex.

3/1 sg masc sg (el) mio 3/2 pl fem sg (las) tuyas 3/3 sg masc pl (los) suyos

"Number" refers to possedee referent. Agreement is established w.r.t this attribute. References to possessor number are made via lexical variation (except when the possessor is third person: "suyo").

As for the attribute "person", these pronominals (coded in ET-ES dictionaries as "nouns") have a person attribute valued as "third" by default since they agree with third person verb forms despite the grammatical person of their referent.

Indefinite: PI These are: "alguien", "nadie", "quienquiera", "qualquiera", "uno", "alguno", "ninguno", "algo", "nada", and "otro". All indefinite determiners can be also used as pronouns. In ET-ES dictionaries this category is split into nouns, nform=pronouns, and the category "quantifier", e.g.: "nada", "nadie", "todos/as".

These are variable in gender and number except for "alguien", "nadie" and "nada".

12.6.2 WH-Type Interrogative

These are: "qu'e, cu'al, qui'en and cu'anto". All of them inflect for number except for "qu'e"; and only "cu'anto/a" inflects for gender and number: "cu'antos/as". In Eurotra dictionaries they are valued as 3rd for person since they agree, when they are the subject of a sentence, with 3rd person verb forms.

Relative These are "que" and "quien". Only "quien" inflects for number. Personal In Spanish the attributes pertinent to personal pronouns are: person, gender (except for dative pronouns), number, case and politeness.

Case is normally related to grammatical functions, thus nominative corresponds to subject, accusative to first object, dative to second object and oblique to those oblique arguments obliRecommendations 166 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R gatorily introduced by a preposition. Note that Spanish has oblique synthetic forms such as "conmigo" (with me).

In the following lines we only give examples for the first person singular forms:

Pers Number Gender Pos Case Funct Pol Infl Ex.

1 sg c nom yo 1 sg c acc me 1 sg c dat me 1 sg c obl mi 1 sg obl conmigo

Reflexive Pronouns

Pers Number Gender Pos Case Funct Pol Infl Ex.

1 sg me 2 sg te 3 sg se 1 pl nos 2 pl os 3 pl se

Reciprocal Pronouns Formally, reciprocal pronouns are the same as reflexive pronouns.

Recommendations 167 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R 12.7 Application to French (Corpus) 12.7.1 Type

Attribute value Example Tag Type dem celle-ci PDETFS

indf quiconque PINDMS poss (*) int/rel lequel PINTMS, PRELMS pers tu PPER2 refl se PREFMS

(*) although possessive pronouns exist in French, they are multiword forms (le n^otre), and therefore are coded as article + noun with the IBMF tagset.

12.7.2 wh-Subtype

Attribute value Example Tag wh-Type int lequel PINTMS

rel qui PRELMS

Note that most interrogative pronouns are also relative pronouns in French. 12.7.3 Person In the IBMF tagset, as in the case of verbs, the person is numbered 1 to 6. The coding of an IBMF verb tag into the EAGLES scheme therefore implies using both the Person and Number information.

Attribute value Example Tag Person 1 je, nous PPER1, PPER4

2 tu, vous PPER2, PPER5 3 il, ils PPER3M, PPER6M

12.7.4 Gender Gender applies to all pronouns and determiners in French, although it can be without a morphological realization, as in the case of the personal pronouns of the 1st and 2nd persons (je, tu, nous, vous).

Attribute value Example Tag Gender masculine celui PDETMS

feminine celle PDETFS

12.7.5 Number Number applies to all pronouns and determiners in French.

Recommendations 168 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R

Attribute value Example Tag Number singular lequel PINTMS

plural lesquels PINTMP

12.7.6 Case Pronouns seem to be the only class where the Case attribute can be used in French. As in the case of Italian, subject, object and oblique personal pronouns can be distinguished. Thus the EAGLES classes nom, acc and obl can be used. The oblique pronoun is not coded specifically in the IBMF tagset.

Attribute value Example Tag Case nom tu PPER2

acc te PPOBMS

As in the case of reflexive personal pronouns, the IBMF tagset does not code the person in the object personal pronoun.

Recommendations 169 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R 12.7.7 EAGLES features not applicable Possessor (the number of) does not apply to French. What could apply, as in other Romance languages, is the person of the possessor. Function and Inflection type do not apply to French. Politeness applies to French (the polite "vous") but is not coded specifically in the tagset.

Recommendations 170 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R 12.8 Application to French (Lexicon) 12.8.1 Type ------------ ----------- ----------- ---- Attribute Value Example Code ------------ ----------- ----------- ---- Type personal je p

demonstrat. celui d indefinite certain i possessive le.mien s interrog. lequel t relative quel r ------------ ----------- ----------- ----

Possessive pronouns are compound forms only ("le mien"). The form "mien" is an adjective.

12.8.2 Person ------------ ----------- ----------- ---- Attribute Value Example Code ------------ ----------- ----------- ---- Person first je,ma 1

second tu,ta 2 third il,sa 3 ------------ ----------- ----------- ----

12.8.3 Gender ------------ ----------- ----------- ---- Attribute Value Example Code ------------ ----------- ----------- ---- Gender masculine cet,il m

feminine cette.elle f neutre ce n ------------ ----------- ----------- ----

12.8.4 Number ------------ ----------- ----------- ---- Attribute Value Example Code ------------ ----------- ----------- ---- Number singular certain s

plural certains p ------------ ----------- ----------- ----

Recommendations 171 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R 12.8.5 Case ------------ ----------- ----------- ---- Attribute Value Example Code ------------ ----------- ----------- ---- Case nominative je n

object me a oblique moi o ------------ ----------- ----------- ----

For the French pronominal system, we could use the following mapping to the EAGLES system: Nominative = subject `il' Accusative = direct object `le' Dative = indirect object `lui' Oblique = other

The category "other" corresponds to the reinforcement of subject or object ("Moi, je le dis"), attribute ("C'est moi"), etc.

However, this solution splits "object" into "direct" and "indirect", and this distintion is valid only for the 3rd person pronouns in French (direct: "le, la, les"; indirect: "lui, leur"). Encoding this distinction would duplicate all other forms (direct: "me, te" etc.; indirect: "me, te" etc.). The following mapping applies readily to French personal pronouns:

subject je, tu, il, elle, nous, vous, ils, elles object me, te, le, la, lui, se, nous, vous, les, leur, se oblique moi, toi, lui, elle, soi, nous, vous, eux, elles, soi

12.8.6 Possessor ------------ ----------- ----------- ---- Attribute Value Example Code ------------ ----------- ----------- ---- Possessor singular mon s

plural nos p ------------ ----------- ----------- ----

12.8.7 Combinations ======== =========================================== Code Example ======== =========================================== Ps1fs-s la.mienne (mienne is not a pronoun) Ps1fs-p la.no?tre Ps1fp-s les.miennes

Recommendations 172 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R Ps1fp-p les.no?tres Ps1ms-s le.mien Ps1ms-p le.no?tre Ps1mp-s les.miens Ps1mp-p les.no?tres

Ps2fs-s la.tienne Ps2fs-p la.vo?tre Ps2fp-s les.tiennes Ps2mp-p les.vo?tres Ps2ms-s le.tien Ps2ms-p le.vo?tre Ps2mp-s les.tiens Ps2fp-p les.vo?tres

Ps3fs-s la.sienne Ps3fs-p la.leur Ps3fp-s les.siennes Ps3fp-p les.leurs Ps3ms-s le.sien Ps3ms-p le.leur Ps3mp-s les.siens Ps3mp-p les.leurs

Pp1-sn- je Pp2-sn- tu Pp3msn- il, on Pp3fsn- elle Pp1-pn- nous Pp2-pn- vous Pp3mpn- ils Pp3fpn- elles

Pp1-sj- me (-moi after imperative) Pp2-sj- te (-toi after imperative) Pp3msj- le, se, lui Pp3fsj- la, se, lui Pp3n-j- en, y Pp1-pj- nous Pp2-pj- vous Pp3mpj- les, se, leur Pp3fpj- les, se, leur

Pp1-so- moi Pp2-so- toi Pp3mso- lui, soi Pp3fso- elle, soi

Recommendations 173 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R Pp1-po- nous Pp2-po- vous Pp3mpo- eux, soi Pp3fpo- elles, soi

Pd-fp-- celles, celles-ci, celles-la' Pd-fs-- celle, celle-ci, celle-la' Pd-mp-- ceux, ceux-ci, ceux-la' Pd-ms-- celui, celui-ci, celui-la' Pd-n--- ce, ceci, cela,

Pi-fp-- quelques-unes, certaines... Pi-fs-- aucune, nulle, certaine... Pi-mp-- quelques-uns, certains... Pi-ms-- aucun, nul, quelqu'un, certain...

Pr-fp-- lesquelles, desquelles, auxquelles, qui, que, quoi, dont, Pr-fs-- laquelle, qui, que, quoi, dont, ou Pr-mp-- lesquels, desquels, auxquels, qui, que, quoi, dont Pr-ms-- lequel, duquel, auquel, qui, que, quoi, dont

Pt----- quoi Pt-fp-- lesquelles, desquelles, auxquelles, qui, que Pt-fs-- laquelle, qui, que Pt-mp-- lesquels, desquels, auxquels, qui Pt-ms-- lequel, duquel, auquel, qui, que ===============================================================

Recommendations 174 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R 12.9 Application to Portuguese 12.9.1 Type ------------ -------------------------- ----------- ---- Attribute Value Example Tag ------------ -------------------------- ----------- ---- Type demonstrative isso

indefinite possessive meu interrogative/relative qual,que personal eu refl me recp excl que ------------ -------------------------- ----------- ----

We assume that possessive applies to Portuguese pronouns. (Nevertheless, in the Portuguese GENELEX demo lexicon, the value possessive was included only at category adjective. This was due to the following assumption: if a candidate pronominal item occurs both as a noun phrase and as a modifier to a full np, it will be categorized as an adjective, presuming that its np-like realization is describable as an instance of nominal ellipsis.)

12.9.2 Person As we explained in the section on Verbs, the Portuguese lexicon adopted the decomposition of the feature person into the features person-deixis and person-conc. These features are also used to encode Portuguese personal pronouns.

12.9.3 Gender ------------ ----------- ----------- ---- Attribute Value Example Tag ------------ ----------- ----------- ---- Gender masculine ele

feminine ela neuter ------------ ----------- ----------- ----

The neuter value is not used in Portuguese.

12.9.4 Number ------------ ----------- ----------- ---- Attribute Value Example Tag ------------ ----------- ----------- ---- Number singular ele

plural eles ------------ ----------- ----------- ----

Recommendations 175 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R 12.9.5 Case ------------ ----------- ----------- ---- Attribute Value Example Tag ------------ ----------- ----------- ---- Case nominative eu

genitive me dative me accusative me oblique mim, comigo ------------ ----------- ----------- ----

Recommendations 176 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R 12.10 Application to Danish The values used in the lexical specification within the tables below are based on the EDEMD; however, minor changes have been made here to adapt a few value names to the EAGLES proposal.

12.10.1 Type

Attribute value Example Tag

Type personal jeg pron pers sg1 nom

reflexive sig pron refl sg3/pl3 reciprocal hinanden pron rec possessive min pron poss sg1 com sg demonstrative denne pron dem com sg indefinite nogen pron indef relative hvem pron rel wh-type hvilken pron wh sg com

12.10.2 Wh-subtype All pronouns beginning with hv- (except `hver' (each)) may function both as relative and as interrogative pronouns. The EDEMD specifies each hv- pronoun as both relative and interrogative types without a common supertype like Wh-subtype.

Attribute value Example Tag Wh-subtype relative hvad pron rel

interrogative hvad pron interr

12.10.3 Gender The feature Gender applies to the following types of pronouns: personal pronoun (only 3rd person singular), possessive pronouns (see below) and to the relative/interrogative hvilken. The only cases were Danish distinguishes the feminine/masculine genders are the personal and possessive pronouns in the third person singular. (In the latter case it refers to the gender (sexus) of the owner.)

Attribute value Example Tag

Gender feminine hun pron pers sg3 fem

masculine han pron pers sg3 mas neuter det pron pers sg3 neut common denne pron dem sg com

Recommendations 177 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R 12.10.4 Number The feature Number applies to demonstrative, personal, possessive and reflexive pronouns and to types which can be used in adjective-like functions, e.g. hvilken.

Attribute value Example Tag

Number singular denne pron dem com sg

plural disse pron dem pl

12.10.5 Case The feature Case applies in different ways to personal pronouns and to the other non-adverbial pronouns. Personal pronouns have two distinct cases: subjective and objective, i.e. nominative and oblique cases. The genitive case is covered by the possessive pronouns. All other non-adverbial pronouns have non-genitive and genitive forms.

Attribute value Example Tag

Case nominative hun pron pers sg3 fem nom

oblique hende pron pers sg3 fem obl non-genitive denne pron dem com sg ngen genitive dennes pron dem com sg gen

12.10.6 Politeness Language-specific feature.

The polite form is `De', used for addressing one or more persons, similar to the personal pronoun of third person plural (`de'), but it is always spelled with a capital D. However, the appropriate reflexive pronoun used in addressing is identical with the oblique case `Dem', thus deviating from the regular form of the third person plural `sig'.

12.10.7 Function The attribute Function is not specified as a lexical property within the EDEMD and no tagset has yet been elaborated for the corpus annotation of this feature.

Attribute value Example Tag

Function nominal hendes (er) pron pers ... nomin

attributive denne (bil) pron dem ... attrib predicative (bilen er) min pron poss ... pred adverbial hvor pron wh-type adverbial

Recommendations 178 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R 12.11 Application to Greek The Greek Morphological Lexicon distinguishes pronouns on the basis of the feature Type (e.g. personal, indefinite, possessive, etc.).

It is noted here that a special category, that of pou, has been created for the codification of the word pou which is highly ambiguous in Greek. In fact, it may function as an uninflected relative pronoun, a subordinating conjunction, an adverb or a particle, while the context does not suffice for absolutely correct disambiguation purposes. Therefore, to avoid multiple tagging that would not be solved, this unique word has been elevated into a category of its own.

In the same way, the special category of enas includes all uses of the ambiguous word enas, one of which is that of the indefinite pronoun. It is also a numeral and the indefinite article in Greek.

In the following paragraphs, we present all values for the types of pronouns and the codification relevant to each type in more detail.

12.11.1 Pronoun Types The following table shows the attribute Type and the values it takes in the Greek Morphological Lexicon.

From the values recognised by the Greek Morphological Lexicon, the majority is foreseen by the Eagles Level 1 codification scheme. These are the values dem, pers, poss, indef and reflex. The distinction between interrogative and relative pronouns is performed at the same level, while three language specific types of pronouns are coded, namely, definite, clitics and relativeindefinite pronouns. More details on each of these types are given in the sections that describe them.

Attribute value ex. tag

Type dem ekeinos PnDm03MaSgNm

poss mou PnPo01CoSgGe indef kapoios PnId03MaSgNm pers eghw PnPe01CoSgNm interr poios PnIr03MaSgNm rel opoios PnRe03MaSgNm reflex (tou eautou tou) PnRf03MaSgGe l-spec relindef osos PnRi03MaSgNm l-spec def (o idhios) PnDf03MaSgNm l-spec clitic me PnCl01CoSgAc

Recommendations 179 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R 12.11.2 Personal Pronouns Personal Pronouns in Greek are morphologically marked for Person, Number and Case; third person pronouns are also marked for Gender.

Two kinds of forms are distinguished : strong and weak pronouns. Under this type, we only include strong forms; weak forms are coded as having type "clitic" (see relevant section below).

Personal example tag Person Number Gender 1 sg common eghw PnPe01CoSgNm 2 sg common esu PnPe02CoSgNm 3 sg masc autos PnPe03MaSgNm 3 sg fem auty PnPe03FeSgNm 3 sg neuter auto PnPe03NeSgNm 1 pl common emeis PnPe01CoPlNm 2 pl common eseis PnPe02CoPlNm 3 pl masc autoi PnPe03MaPlNm 3 pl fem autes PnPe03FePlNm 3 pl neuter auta PnPe03NePlNm

In the above table, we have given personal pronouns only in the nominative case. All of them are inflected in the genitive case (emena, esena, autou, autys, autou, emas, esas, autwn, autwn, autwn), and the accusative (emena, esena, auton, auty(n), auto, emas, esas, autous, autes, auta). Only two forms appear in the vocative case, namely the second person singular (esu) and plural (eseis).

12.11.3 Politeness form In Greek, two forms of the second person of the personal pronoun are used when addressing someone:

- the second singular (esu) in case of familiarity between addresser and addressee, and - the second plural (eseis) for politeness form. The pronouns naturally present agreement as regards number with the verb. As regards, however, participial constructions and adjectives, agreement with verb is overruled by agreement with the sex and number of the addressee(s):

Eseis eiste haroumeny me ta nea; - fem., sing. The feature of Politeness does not correspond to a specific attribute in the ILSP Morphological

Recommendations 180 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R Lexicon. 12.11.4 Reflexive Pronoun In Greek, the reflexive pronoun is formed periphrastically by the noun "eautos" preceded by the definite article and followed by the possessive pronoun (e.g. tou eautou mou). It is inflected only for two cases, genitive (eautou) and accusative (eauto).

In the Greek Morphological Lexicon, we have coded the relevant forms of the lemma "eautos" as a noun; when encountered in the above compound form, it must be recognised as a reflexive pronoun. The following table shows examples for the values it must be assigned when found in the genitive case.

Reflexive example tag Person Number Case 1 sg gen (tou eautou mou) (PnRf01CoSgGe) 2 sg gen (tou eautou sou) (PnRf02CoSgGe) 3 sg gen (tou eautou tou) (PnRf03MaSgGe) 3 sg gen (tou eautou tys) (PnRf03FeSgGe) 3 sg gen (tou eautou tou) (PnRf03NeSgGe) 1 pl gen (twn eautwn mas) (PnRf01CoPlGe) 2 pl gen (twn eautwn sas) (PnRf02CoPlGe) 3 pl gen (twn eautwn tous) (PnRf03CoPlGe)

12.11.5 Possessive pronouns The weak forms of the genitive case of personal pronouns serve as possessive pronouns for Greek. They are distinguished according to the Person they refer to and code the Number of the possessor (one or more). The third person in the singular number also distinguishes the Gender of the possessor, while in the plural number, a common form covers all three genders.

Possessive Person Number Gender example tag 1 sg common mou PnPo01CoSgGe 1 pl common mas PnPo01CoPlGe 2 sg common sou PnPo02CoSgGe 2 pl common sas PnPo02CoPlGe 3 sg masc tou PnPo03MaSgGe 3 sg fem tys PnPo03FeSgGe 3 sg neut tou PnPo03NeSgGe 3 pl common tous PnPo03CoPlGe

In certain cases, most usually for emphasis, the possessive pronoun is formed periphrastically on the basis of the above pronoun preceded by the adjective "dhikos" which must agree in

Recommendations 181 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R gender, number and case with the possessed objects:

ta dhika mas vivlia - neuter, pl., nom/acc

oi dhikes mas eikones - fem., pl., nom

Information on the possesed object (e.g. gender) is not represented in the possessive pronouns, and, thus, is not coded by an attribute in the Morphological Lexicon.

12.11.6 Demonstrative pronouns Five demonstrative pronouns are recognised in Greek, all of which are inflected for gender, number and case agreeing with the referred object. The following table offers examples for all of them in the singular nominative case of the masculine form.

Demonstrative example tag

autos PnDm03MaSgNm toutos PnDm03MaSgNm ekeinos PnDm03MaSgNm tetoios PnDm03MaSgNm tosos PnDm03MaSgNm

The first three of the above demonstratives may occur either on their own or followed by the nouns to which they refer, in which case the nouns are preceded by an article. The last two serve only as modifiers of nouns and do not have an independent nominal status.

The type of deixis is not encoded, although it is implicit in each pronoun. One of the pronouns, toutos, is more common in everyday use in spoken language, rather than in the written/formal usage of Greek.

Recommendations 182 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R 12.11.7 Indefinite pronouns

Indefinite example tag

kanenas PnId03MaSgNm kapoios PnId03MaSgNm merikoi PnId03MaPlNm kati PnId03CoNvIc katiti PnId03NeNvIc tipote PnId03CoNvIc kamposos PnId03MaSgNm kathe PnId03CoNvIc kathenas PnId03MaSgNm katheti PnId03NeSgIc allos PnId03MaSgNm

The first indefinite pronoun, kanenas, does not form a plural and has two forms for the masculine gender:

kanenas and kaneis Both kanenas and tipote have a negative as well as an indefinite meaning:

kanenas = nobody or somebody

tipote = nothing or something

They have only the indefinite interpretation in affirmative and interrogative sentences, and both interpretations in negative sentences and answers to interrogatives:

Eheis tipote;

but Dhen ehei tipote pou na tou pyghainei

The two above pronouns, along with kati, katiti, tipote, kathe and katheti are uninflected. The pronoun merikoi, on the other hand, does not form the singular, but inflects for case.

All other pronouns are inflected as adjectives (i.e. they mark gender, number and case). Most of them function both as pronouns and pronominal adjectives. The pronoun kathe functions only as a pronominal adjective; its "respective" pronoun is kathenas.

12.11.8 Interrogative Pronouns Three interrogative pronouns are distinguished for Greek.

Recommendations 183 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R

Interrogative example tag

ti PnIr03CoNvIc poios PnIr03MaSgNm posos PnIr03MaSgNm

As indicated by the values of its tag, the first pronoun, ti, is indeclinable, while the other two are inflected for gender, number and case, agreeing with the object they refer to. They are all used both as pronouns and pronominal adjectives.

12.11.9 Relative pronouns Two relative pronouns exist for Greek, one of which, namely the indeclinable pou, is encoded in the Morphological Lexicon as a separate category. The other pronoun is opoios, which is always preceded by the definite article and agrees in gender, number and case with its referent; it is considered more "formal" than "pou". Again, in the following table, only the nominative case of the singular of the masculine is given.

Relative example tag

opoios PnRe03MaSgNm pou Po

12.11.10 Relative-Indefinite Pronouns

Relative-Indefinite example tag

osos PnRi03MaSgNm o,ti PnRi03CoNvIc

Contrary to o,ti, which is indeclinable, osos is inflected for gender, number and case. 12.11.11 Definite Pronouns

Definite example tag

(o idhios) (PnDf03MaSgNm) (monos mou) (PnDf01MaSgNm)

Recommendations 184 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R The definite pronouns are actually adjectives which function as pronouns when the following conditions are fulfilled:

- idhios must be preceded by the definite article, and it must follow or it must precede the noun it refers to, which has to have the definite article:

Eghw y idhia tha to kanw or O idhios o Takys tha to kanei

- monos must appear without an article and must be followed by one of the possessive pronouns:

Efughe monos tou Therefore, they are coded in the Morphological lexicon as adjectives, and their recognition as Pronouns depends on the use of a multi-word tagger.

12.11.12 Clitic Pronouns These are actually the weak forms of the personal pronouns. They are morphologically marked for Person, Number and Case. For certain cases, there exists no weak form, as observed in the following table.

Recommendations 185 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R

Personal example tag Person Number Case Gender 1 sg gen common mou PnCl01CoSgGe 1 sg acc common me PnCl01CoSgAc 2 sg gen common sou PnCl02CoSgGe 2 sg acc common se PnCl02CoSgAc 3 sg nom masc tos PnCl03MaSgNm 3 sg gen masc tou PnCl03MaSgGe 3 sg acc masc ton PnCl03MaSgAc 3 sg nom fem ty PnCl03FeSgNm 3 sg gen fem tys PnCl03FeSgGe 3 sg acc fem ty(n) PnCl03FeSgAc 3 sg nom neuter to PnCl03NeSgNm 3 sg gen neuter tou PnCl03NeSgGe 3 sg acc neuter to PnCl03NeSgAc 1 pl gen common mas PnCl01CoPlGe 1 pl acc common mas PnCl01CoPlAc 2 pl gen common sas PnCl02CoPlGe 2 pl acc common sas PnCl02CoPlAc 3 pl nom masc toi PnCl03MaPlNm 3 pl gen masc tous PnCl03MaPlGe 3 pl acc masc tous PnCl03MaPlAc 3 pl nom fem tes PnCl03FePlNm 3 pl gen fem tous PnCl03FePlGe 3 pl acc fem tis/tes PnCl03FePlAc 3 pl nom neuter ta PnCl03NePlNm 3 pl gen neuter tous PnCl03NePlGe 3 pl acc neuter ta PnCl03NePlAc

Recommendations 186 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R 12.12 Application to Swedish The following features are relevant for Swedish pronouns; Type, Number, Person, Gender and Case.

Type Number Person Gender Case personal singular first masculine nominative reflexive plural second feminine oblique reciprocal third inanimate possesive ordinary demonstrative nonneuter genitive (determinative) indefinite neuter relative interrogative

12.12.1 Type This is a lexically determined feature, normally prescribing the syntactic use of the subcategory. The values (subcategories) are personal, reflexive, reciprocal, possessive, demonstrative, determinative, indefinite, relative and interrogative. (Most of the subcategories may function as both a kind of adjectives and a kind of nouns).

Attribute value Example Tag Type personal jag pron-pers-sg1-nom

reflexiv mig pron-refl-sg1 reciprocal varandra pron-rec possessive min pron-poss-sg1-com-sg demonstrative denna pron-dem-com-sg indefinite n*agon pron-indef relative vilken pron-rel interrogative vem pron-wh-sg-com

12.12.2 Number The values are singular and plural. The feature is applicable to most of the subcategories.

Attribute value Example Tag Number singular denna pron-dem-com-sg

plural dessa pron-dem-pl

12.12.3 Person The values are first, second and third. The feature is applicable to personal and possessive pronouns. The only independent reflexive pronoun, sig, is third Person, singular and plural. There also exists a "possessive reflexive" sin/sitt, third Person singular.

Recommendations 187 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R

Attribute value Example Tag Person first jag pron-pers-sg1

second du pron-pers-sg2 third hon pron-pers-sg3

12.12.4 Gender The values are masculine, feminine, inanimate, nonneuter and neuter. The value nonneuter comprises the first three values when opposed to neuter. (The complications are due to the interaction between a grammatical two-gender system and a semantic anaforic system for animateness and sex.) The feature is applicable in singular for personal and possessive pronouns and, in the two-gender form, for most other subcategories.

Attribute value Example Tag Gender feminine hon pron-pers-sg3-fem

masculine han pron-pers-sg3-mask inanimate den pron-pers-sg3-in neuter det pron-pers-sg3-neu

Attribute value Example Tag Gender neuter vilket pron-rel-neu

nonneuter vilken pron-rel-noneu

12.12.5 Case There are two oppostions in case relevant for pronouns: nominative vs oblique for personal pronouns and ordinary vs genitive for most other subcategories. The value ordinary comprises in principle nominative and oblique. (The possessive pronouns could be regarded as the genitives of the personal pronouns, which would complete the case system, but that would be contrary to the grammatic traditions.)

Attribute value Example Tag Case nominative hon pron-pers-sg3-fem-nom

oblique henne pron-pers-sg3-fem-obl non-genitive denne pron-dem-com-sg-ngen genitiv dennes pron-dem-com-sg-gen

Recommendations 188 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R 12.13 Application to Catalan The Catalan corpus and lexicon does not subtype pronouns, but we could adapt some informations automatically to the EAGLES recommendations: person, gender, number. Changes as Type tags could be applied by hand in the lexicon, because there is a small number of lemmas involved in.

Recommendations 189 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R 12.14 Application to Irish With the possessive classified as a Determiner, and the function of the Relative Pronoun subsumed in the complementizer system, the schema for Pronouns is quite simple. The contrastive form, already presented for nouns, is specific to Irish. Wh-forms can be treated as forms of determiners.

12.14.1 Type and Wh-type =========================================================================== Attribute Value Example Tag =========================================================================== Type Personal me', tu', se', si', e'/i' [I, you, he, she, it] PR

Contrastive mise, tusa, eisean, ise [I, you, he, she] PRCON Reflexive me' fe'in, tu' fe'in [I myself, you yourself] PRFEIN ===========================================================================

It would be possible to add Relative value to cover the usages of the complementizers a+L, a+E and go (See Conjunction) that have a relative function, and there may be arguments for doing this later in multilingual applications.

12.14.2 Person =========================================================================== Attribute Value Example Tag =========================================================================== Person 1 me', mise, me' fe'in; muid, muidne PR1s/p/CON

2 tu', tusa, tu' fe'in; sibh, sibhse PR2s/p/CON 3 se'/si', e'/i'; eisean/ise; iad, siad, iadsan PR3s/p/m/f/ ===========================================================================

12.14.3 Gender ======================================================== Attribute Value Example Tag ======================================================== Gender Masculine se', e' [he, it (masc.)] PR3sm

Feminine si', i' [she, it (fem.)] PR3sf ========================================================

12.14.4 Number ======================================================================== Attribute Value Example Tag ======================================================================== Number Singular Me', tu', se', e', i' [he, it (masc./fem.)] PR1/2/3/s

Plural Muid, sibh, siad [we, you (pl.), they] PR1/2/3/p ========================================================================

Recommendations 190 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R 12.14.5 EAGLES features that do not apply Case, Possessor, Politeness and Functionality do not apply.

Recommendations 191 May, 1996 EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R 13 Determiner

P Type whType Pers Gend Numb Case Pos Infl M 1 m s

U 2 f p

L 3 n T G dem 1 m sg sg

E int 2 f pl pl N poss 3 n

E card

L indf E def X rel

part excl A 1 m s

l 2 f p D 3 n

N poss 1 m s nom

E dem 2 f p gen R indf int 3 mf sp dat

C int/rel rel acc

L dem int 1 m s nom sg

e indf rel 2 f p gen pl e poss 3 n dat c int/rel c acc

E-L0 DETERMINER

E dem 1 m sg sg

indf 2 f pl pl L poss 3 n

1 int

rel L int/rel nom

2 dem/indf gen a dat

acc 2 Fr/It excl It c It n Ge infl b Ge no-infl

Recommendations 192 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R 13.1 Comments One of the major problems presented by this category is that the Romance tradition makes use of the label Pronominal Adjective, whereas other languages (English) use Determiner. The two classes are not mappable one on to the other. This is a crucial problem, involving the different behaviour of the Determiners and Pronominal Adjectives. The choice of calling e.g. some and all respectively Determiner and Predeterminer in English (both Indefinite Pronominal Adjectives in the Italian tradition) rests on their particular distribution in context: some children, all the children, and this can also work in Italian: alcuni ragazzi, tutti i ragazzi. However, if calling Determiners the possessives in English works for their complementary distribution wrt the article (my book, the book), this is not the case in Italian, because il libro does not have correspondence in *mio libro. (As far as Italian possessives are concerned, a complementary distribution with article is found only with a closed number of family nouns: mio/il padre, mia/la madre, mio/il fratello, mia/la sorella). The particular behaviour of possessives perhaps influenced the GENELEX choice to opt for a separate Determiner category, as well as the Adjective (where Pronominal Adjectives are also included) and Pronoun categories. The situation in the GENELEX model is as follows:

le n^otre PRON n^otre chien DET le chien est n^otre ADJ

This fact implies the presence of the values typical of Pronominal Adjective in the table of the Adjective category, among the values of the feature Type.

In NERC, the problem of the use of different tags, i.e. Determiner and Pronominal Adjective, in different traditions has been raised, and the proposal has been to opt for the label Determiner, without any further functional distinction. The TEI work-group on annotation (TEI AI1W2 1991) has made the choice of inserting Determiners in the Adjective category, with the feature `pronominal'.

Recommendations 193 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R 13.2 Application to Italian In Italian, both corpus and dictionary distinguish Determiners according to the feature Type (e.g. indefinite, demonstrative, possessive, etc.).

In the following sections, each Type will be discussed in detail. 13.2.1 Type and wh-Type

Attribute value ex. tag

Type dem questo (libro) DD/ms

poss mio (padre) DP/ms indf ogni (uomo) DI/ms wt-Type int quale (domanda) DT/ns

rel il quale (uomo) DR/ms excl quanto (sole)! DE/ms

13.2.2 Possessive

Possessive example It.tag

(i) miei (amici) DP/mp mio (cugino) DP/ms (la) nostra (casa) DP/fs (i) nostri (cugini) DP/mp

In English, Possessives are classified as Determiners on the basis of their complementary distribution wrt the article (my book, the book). This kind of distribution in Italian works only with a closed number of family nouns, used in the singular: mio/il padre, mia/la madre, mio/il fratello, mia/la sorella (see the table above). Other nouns do not show this correspondence, e.g il libro / *mio libro. At the level of encoding, this different behaviour is not represented. Possessives are inflected for Number and Gender and agree with the noun; they are distinguished according to the Person which is referred to (see the table for Pronouns above).

Scrivo con la tua DP/fs penna, perche' non ho la mia The two Italian possessives: altrui (of other people) and proprio (own) can be used in determiner function:

spende il denaro altrui DP/nn, non il proprio occorre dare del proprio DP/ms denaro, non dell'altrui

Recommendations 194 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R As already mentioned for Possessive Pronouns, information about the possessor is not encoded in the Italian lexicon or corpus, but it can be inferred from the lemma.

13.2.3 Demonstrative

Demonstrative example It.tag

questo (uomo) DD/ms quelle (donne) DD/fp

13.2.4 Indefinite

Indefinite example It.tag

ogni (uomo) DI/ms alcune (donne) DI/fp

Italian Indefinites are inflected for Gender and Number. They cover the English class of Quantifiers; some English practices (Brown) usually encode some quantifiers as Predeterminers, (e.g. all the girls), on the basis of their peculiar behaviour, i.e. they precede the determiner. Tutto in Italian has the same behaviour, which is not annotated:

tutte le ragazze DI/fp Some can only be pronominal adjectives, i.e. determiners in this proposal: ogni, qualche, qualunque, qualsiasi, qualsivoglia.

ogni colpa si sconta DI/fs Alcuno, ciascuno, taluno, nessuno, tutto, alquanto, poco, molto, troppo, tanto can have both the pronoun and pronominal adjective/determiner function:

ho letto tutto il libro DI/ms 13.2.5 Interrogative

Interrogative example It.tag

quanto (zucchero)? DT/ms quale (libro)? DT/ns

Recommendations 195 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R Interrogatives are inflected for Gender and Number. Che, quale, quanto can be either pronouns or pronominal adjectives/determiners:

Quale vestito scegli? DT/ns

13.2.6 Exclamatory

Exclamatory example It.tag

quanto (vento)! DE/ms che (orrore)! DE/nn

Che, quale, quanto can also have exclamatory value:

quante persone hanno aderito! DE/fp

13.2.7 Relatives

Relative example It.tag

il quale DR/ms

Il quale is inflected for Gender and Number, and can be used in adjectival function:

..., il quale film mi e' piaciuto molto DR/ms It should be noted that it constitutes a bigram and should be encoded as a multiword expression.

Recommendations 196 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R 13.3 Application to German 13.3.1 Type

Attribute value example tag

Type demonstrative dieses (Buch) PDEMAT:Neut.Nom.Sg

indefinite irgendein (Buch) PROAT:Neut.Nom.Sg possessive mein (Buch) PPOSAT:Neut.Nom.Sg interrogative welches (Buch) ? PWAT:Neut.Nom.Sg

13.3.2 Gender

Attribute value example tag

Gender masculine dieser (Mann) PDEMAT:3.Sg.Nom.Masc

feminine meine (Mutter) PPOSAT:Fem.Nom.Sg neuter welches (Buch) PWAT:Neut.Nom.Sg

13.3.3 Number

Attribute value example tag

Number singular kein (Mensch) PROAT:Masc.Nom.Sg

plural welche (Kinder) PWAT:Neut.Nom.Pl

13.3.4 Case

Attribute value example tag

Case nominative kein (Mensch) PROAT:Masc.Nom.Sg

genitive dieses (Mannes) PDEMAT:Masc.Gen.Sg dative keinem (Menschen) PROAT:Masc.Dat.Sg accusative welchen (Mann)? PWS:Masc.Akk.Sg

13.3.5 Possessor The feature possessor is not encoded in the IMS-Tagset. It can be determined from the lemma.

Recommendations 197 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R 13.4 Application to English 13.4.1 Type

Attribute values Examples Tags Type possessive your DV2

demonstrative those DDp wh-type what DW indefinite every DIs

13.4.2 Person

Attribute values Examples Tags Person first my, our DV1ps, DV1pp

second your DV2 third his, her, their DV3psM, DV3psF, DV3ppN

For determiners, as for pronouns, distinctions of person, gender and number sometimes apply. Person applies to possessive determiners, and gender applies to third person singular possessive determiners.

13.4.3 Gender

Attribute values Examples Tags Gender masculine his DV3psM

feminine her DV3psF

13.4.4 Number Number applies to demonstrative determiners and, in some cases, to indefinite determiners:

Attribute values Examples Tags Number singular this, much DDs, DIs

plural these, many DDp, DIp

The number of the possessor is also distinguishable for possessive determiners. However, for the present tagset, this attribute is omitted, being unimportant as an indicator of syntactic function.

Recommendations 198 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R 13.5 Application to Spanish Case does not apply to Spanish determiners. As for the other features, each Type is discussed in detail in the following sections.

Demonstrative: DD

Pers Number Gender Pos Case Ex.

sg masc este (libro) pl fem esas (casas) sg masc aquel (chico)

Only Number and Gender are pertinent attributes for Determiner demonstratives. These determiners reflect deictic degree of remoteness, which is not, however, coded in ET-ES dictionaries.

Possessive: DP *Possessive Determiners need the same attributes as the Pronominal Possessives. They differ in that in this case we have determiners valued as common wrt gender and possessor:

Pers Number Gender Pos Case Ex.

1 sg c sg mi (libro) 2 sg fem pl vuestra (casa) 3 sg c c su (libro)

In this case, the attribute "person" refers to semantic person. This attribute is not coded in ET-ES dictionaries.

Indefinite DI The main ones are: alg'un, ning'un, cierto, varios, cualquier, poco, bastante, mucho, demasiado, dem'as, cuanto, todo, cada, tal,otro, un, tanto. They are variable in gender except for bastante, cualquiera, cada, dem'as and tal, and variable in number except for cada, demas and varios.

Interrogative These are: qu'e (libros), cu'al (caja), and cu'anto (dinero). "Qu'e is invariant in gender and number. Only cu'anto inflects for gender.

Relative The Spanish relative determiner inflects for number and gender (cuyo, cuya, cuyos and cuyas).

Recommendations 199 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R 13.6 Application to French (Corpus) 13.6.1 Determiner type

Attribute value Example Tag Type dem,poss cette,ma (maison) DETRFS

indf aucune (maison) ADJIFS int quelle (maison) DINTFS

Note that the IBMF tagset only distinguishes two determiner types: interrogative and others (demonstrative, possessive, and articles). The indefinite determiner is classified among the adjectives in the said tagset.

Recommendations 200 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R 13.7 Application to French (Lexicon) 13.7.1 Det.-Type ------------ ----------- ----------- ---- Attribute Value Example Code ------------ ----------- ----------- ---- Type demonstrat. ce d

indefinite certain i possessive mon p interrogat. quel t ------------ ----------- ----------- ----

13.7.2 Person ------------ ----------- ----------- ---- Attribute Value Example Code ------------ ----------- ----------- ---- Person first ma 1

second ta 2 third sa 3 ------------ ----------- ----------- ----

13.7.3 Gender ------------ ----------- ----------- ---- Attribute Value Example Code ------------ ----------- ----------- ---- Gender masculine cet,il m

feminine cette,elle f ------------ ----------- ----------- ----

13.7.4 Number ------------ ----------- ----------- ---- Attribute Value Example Code ------------ ----------- ----------- ---- Number singular certain s

plural certains p ------------ ----------- ----------- ----

13.7.5 Possessor ------------ ----------- ----------- ---- Attribute Value Example Code ------------ ----------- ----------- ---- Possessor singular mon s

plural nos p ------------ ----------- ----------- ----

Recommendations 201 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R 13.7.6 Combinations ========= ============================================= Tag Example ========= ============================================= Ds1fss-- ma (tasse) Ds1fsp-- notre (tasse) Ds1fps-- mes (tasses) Ds1fpp-- nos (tasses) Ds1mss-- mon (livre) Ds1msp-- notre (livre) Ds1mps-- mes (livres) Ds1mpp-- nos (livres)

Ds2fss-- ta (tasse) Ds2fsp-- votre (tasse) Ds2fps-- tes (tasses) Ds2fpp-- vos (tasses) Ds2mss-- ton (livre) Ds2msp-- votre (livre) Ds2mps-- tes (livres) Ds2mpp-- vos (livres)

Ds3fss-- sa (tasse) Ds3fsp-- leur (tasse) Ds3fps-- ses (tasses) Ds3fpp-- leurs (tasses) Ds3mss-- son (livre) Ds3msp-- leur (livre) Ds3mps-- ses (livres) Ds3mpp-- leurs (livres)

Dd-fs--- cette Dd-ms--- cet, ce Dd-fp--- ces Dd-mp--- ces

Dn-fs--- aucune, nulle, certaine, toute, chacune... Dn-ms--- aucun, nul, certain, tout, chacun... Dn-fp--- certaines, toutes... Dn-mp--- certains, tous...

Dt-fs--- quelle Dt-ms--- quel Dt-fp--- quelles Dt-mp--- quels Di-fs--- aucune, nulle, certaine, toute, chacune...

Recommendations 202 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R Di-ms--- aucun, nul, certain, tout, chacun... Di-fp--- certaines, toutes... Di-mp--- certains, tous... ========= ======= =============================================

Recommendations 203 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R 13.8 Application to Portuguese 13.8.1 Type As is easily observable, articles are included under the category determiner in the Portuguese application:

------------ ---------------------------------- ----------- ---- Attribute Value Example Tag ------------ ---------------------------------- ----------- ---- Type demonstrative

indefinite um possessive interrogative/relative =============================================== =========== ==== l-specif definite o

quantifier algum cardinal um ----------------------------------------------- ----------- ----

13.8.2 Wh-Type Not used in the Portuguese application.

13.8.3 Person Not used in the Portuguese application.

13.8.4 Gender ------------ ----------- ----------- ---- Attribute Value Example Code ------------ ----------- ----------- ---- Gender masculine um

feminine uma neuter ------------ ----------- ----------- ----

13.8.5 Number ------------ ----------- ----------- ---- Attribute Value Example Code ------------ ----------- ----------- ---- Number singular algum

plural alguns ------------ ----------- ----------- ----

13.8.6 Case Not used in the Portuguese application

Recommendations 204 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R 13.8.7 Possessor Not used in the Portuguese application.

Recommendations 205 May, 1996 EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R 13.9 Application to Danish The label Determiner as used within the present EAGLES proposal obviously covers a number of lexical items classified as pronouns and quantifiers. Traditional Danish grammars and dictionaries do not operate with a common designation for the determination function. The application of the label Determiner as a member of the Danish tagset for corpus annotation has not yet been clarified. In Danish, from a functional point of view, members of certain word classes, like pronoun or noun (viz. subclasses specifier and classifier), can be considered as determiners:

13.9.1 Determiner type

Attribute value Example Tag

Type demonstrative denne pron dem com sg

possessive min pron poss sg1 com sg quantifier begge quant ordinal tredie ord cardinal tre card specifier halvdelen (af) spec classifier en gruppe class

Although articles and possessive genitives also function as determiners, they have not been included in the table above. We follow here an approach wherein articles (as a separate category) are distinguished from determiners, and possessive genitives do not constitute a category as such.

Recommendations 206 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R 13.10 Application to Greek For the classification of the items that fall under the categories of Pronouns and Determiners, the ILSP Morphological Lexicon followed the traditional grammar distinctions. Although, as mentioned in the relevant section, "absolute" pronouns can be considered only the personal and demonstrative pronouns, we have included (in a broad interpretation of the term) all adjectival pronouns as well.

Furthermore, it is the case that certain "indefinite pronouns" have a strong quantificational character; such cases are "arketoi" (several), "merikoi" or "kamposoi" (some), "kanenas" (none). Based on their distributional behaviour, we could classify them as Determiners: they appear right before the Noun, or alone, they are never modified by degree adverbials, they are never introduced by articles, etc. However, we opted for the traditional classification of all these items under the category of Pronouns. Thus, since there is no attribute available for their codification (no tag assigned), no tables are presented in this section.

Recommendations 207 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R 13.11 Application to Swedish This separate category is, as such, irrelevant for Swedish since it is covered by pronouns, numerals and (free) articles.

Recommendations 208 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R 13.12 Application to Catalan Determiner is not a category in Catalan corpus and lexicon. Determiners are tagged as pronouns, adjectives, or both, depending on the case. But if we take into account the number of lemmas involved in, we should conclude that an adaptation to the EAGLES recommendations should be easy.

Recommendations 209 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R 13.13 Application to Irish This is a broad category because of the decision (which is reversible) to include possessive pronouns, demonstratives, numbers and quantifiers. The article could be included, but may also be treated separately, as recommended.

13.13.1 Type ======================================================================== Attribute Value Example Tag ======================================================================== Type Possessive mo, do, a, a'r, bhur, a [my, your, etc.] DP

Demonstrat (an leabhar) seo [this (book)] DD Numeral tri' (leabhar) [three (books)] DN Quantifier (leabhar) e'igin [some (book)] DQ Ordinal (an) ceathru' (leabhar) [(the) fourth (book)] DO Contextual (an leabhar) ce'anna/eile [(the) same/other (book)] DC Wh-forms ce', ca', cad DW ========================================================================

13.13.2 Person, Gender, Number, Case and Possessor Only the Possessive has the attributes Person, Gender and Number. The gender distinction is present only in the third person singular, marked by the presence or absence of an initial mutation (lenition) on the noun it governs. Number in the third person is also marked only by an initial mutation (eclipsis). So we have a (her), a+L (his), and a+E (their). The attributes Case and Possessor do not apply.

Recommendations 210 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R 14 Article

ARTICLE Type Gen Num Case MULTILEX def m s nom

indf f p gen

n dat

acc voc GENELEX (*in Det)

def m s AlethDic indf f p

n def m s nom NERC indf f p gen

mf sp dat

acc def m s nom indf f p gen Leech n dat

c acc

EAG-L0 ARTICLE

def m sg EAG-L1 indf f pl

n

nom EAG-L2a gen

dat acc EAG-L2b Fr part It c

Recommendations 211 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R 14.1 Comments Articles constitute a separate category, as is found in many lexicons and annotation practices, even though, as already pointed out in section 5, in others (e.g. Penn) they are found incorporated into the Determiner category.

The separation of the two categories is not a problem, since, as Articles form a very restricted closed class, they can also easily be distinguished from Determiners in the latter cases, and no serious mapping problem arises.

Recommendations 212 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R 14.2 Application to Italian Article traditionally constitutes a class of its own.

14.2.1 Article The class of Articles has a finite list of members. The two possible Types `definite' and `indefinite' are marked in the lexicon (RD and RI respectively), while in the corpus they are not encoded. The other pertinent features are Number and Gender.

Article It.tag

Type Gender Number definite indefinite m s il, lo un, uno RD/ms n s l' RD/ns m p i, gli RD/mp f s la una/un' RI/fs f p le RI/fp

Un, il are graphical variants of uno, lo, used before words beginning with simple consonant, or beginning with a consonantal group plus l, r : un trono, il cloro (not if the the group of consonants begins with s).

Uno, lo are used before the consonantal digram beginning with s, before the digram gn and before z, x and consonantic groups whose second element is not l, r. They are also used before words beginning with a vowel and in this case they are elided.

The article helps in general in disambiguating other words in the context: l', sometimes, is not sufficient to disambiguate: l'insegnante capace parte domani (can be both masculine and feminine).

Recommendations 213 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R 14.3 Application to German 14.3.1 Types

Attribute value example tag

Type definite der (Mann) ART:Def.Masc.Nom.Sg

indefinite ein (Mann) ART:Indef.Masc.Nom.Sg

14.3.2 Gender

Attribute value example tag

Gender masculine der (Mann) ART:Def.Masc.Nom.Sg

feminine die (Mutter) ART:Def.Fem.Nom.Sg neuter ein (Buch) ART:Indef.Neut.Nom.Sg

14.3.3 Number

Attribute value example tag

Number singular der (Mann) ART:Def.Masc.Nom.Sg

plural die (Dinge) ART:Def.Neut.Nom.Pl

14.3.4 Case

Attribute value example tag

Case nominative der (Mann) ART:Def.Masc.Nom.Sg

genitive des (Vaters) ART:Def.Masc.Gen.Sg dative einem (Mann) ART:Indef.Masc.Dat.Sg accusative den (Mann) ART:Indef.Masc.Akk.Sg

Recommendations 214 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R 14.4 Application to English The attributes of Gender and Case do not apply to articles in English. Only two words belong to this category: the and a, (an being the variant of a occurring before vowels.) The is invariable.

14.4.1 Type

Attribute values Examples Tags Type definite the ATD

indefinite a, an ATIs

14.4.2 Number

Attribute values Examples Tags Number singular a, an ATIs

plural - -

Recommendations 215 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R 14.5 Application to Spanish 14.5.1 Article Type

Attribute value Example Tag

Type definite el msdefs

indfefinite un msindef

14.5.2 Gender

Attribute value Example Tag

Gender m el

f la n lo

14.5.3 Number

Attribute value Example Tag

Number s un

p unos

There are different positions on the "neuter" article in the Spanish grammatical literature. It is only identifible in singular, since in the plural the form would coincide with the masculine plural. Normally it is used for nominalizations:

lo bueno es ... (the good thing is...) 14.5.4 Case Case does not apply to Spanish articles.

Recommendations 216 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R 14.6 Application to French (Corpus) 14.6.1 Definiteness French has definite and indefinite articles (le, la les, un, une, des). However, there is no distinction in the IBMF tagset because it would have no predictive power and it can be retrieved from the graphic form. In the tagset, articles are coded as regular determiners.

14.6.2 Gender

Attribute value Example Tag Gender masculine le, un DETRMS

feminine la, une DETRFS

14.6.3 Number

Attribute value Example Tag Number singular le, un DETRMS

plural les, des DETRMP

14.6.4 EAGLES features not applicable Case does not apply to French.

14.7 Application to French (Lexicon) 14.7.1 Type ------------ ----------- ----------- ---- Attribute Value Example Code ------------ ----------- ----------- ---- Type definite le d

indefinite un i ------------ ----------- ----------- ----

14.7.2 Gender ------------ ----------- ----------- ---- Attribute Value Example Code ------------ ----------- ----------- ---- Gender masculine le m

feminine la f ------------ ----------- ----------- ----

14.7.3 Number ------------ ----------- ----------- ---- Attribute Value Example Code ------------ ----------- ----------- ----

Recommendations 217 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R Number singular le s

plural les p ------------ ----------- ----------- ----

14.7.4 Case Case is not applicable to French.

14.7.5 Combinations --------- ----------- Tag Example --------- ----------- Tdms- le Tdfs- la Tdmp- les Tdfp- les Tims- un Timp- une Tifs- des Tifp- des --------- -----------

Recommendations 218 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R 14.8 Application to Danish Gender, number and definiteness are pertinent to Danish articles. The table below covers the whole set of articles. The corpus tagger for Danish will treat articles as incorporated into the pronoun category.

14.8.1 Type

Attribute value Example Tag

Type indefinite en art com sg indef

indefinite et art neut sg indef definite den art com sg def definite det art neut sg def definite de art pl def

14.8.2 Language-specific property The enclitic article

Definiteness of nouns must be expressed in certain cases by means of enclitic articles which function as suffixes: -(en) singular common, -(e)t singular neuter and -(e)ne in plural. In corpus tagging this phenomenon will be annotated as a value of the noun definiteness feature.

Recommendations 219 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R 14.9 Application to Greek Two types of articles are distinguished in Greek: definite and indefinite, each class including one member: - definite article: o, y, to, - indefinite article: enas, mia, ena.

In the first version of the ILSP Morphological Lexicon and corresponding Tagset, these two were both coded under the grammatical category of article, further coded for the feature Type with the values definite and indefinite. At a later stage, however, this was abandoned and currently only the definite article is characterised as article. The indefinite article has been elevated to a category of its own, enas, due to its high ambiguity: besides indefinite article, it can also be a cardinal and an indefinite pronoun. The resolution of the ambiguity, even with the use of the linguistic context, is extremely difficult to perform automatically:

Mono enas anthrwpos ytan ekei. - card.

Ton eidha mia mera. - ind. art. Perase enas apo edhw kai se zytaghe. - ind. pron.

The following tables present the values and tags of the attribute Type as used in the first version, in accordance to the EAGLES proposal.

Attribute value Gr. example Gr. tag

Type definite o AtDf

indefinite enas (AtId)

current codif. : En

Articles are further coded for Gender, Number and Case. 14.9.1 Gender, Number and Case Articles behave as adjectives in the sense that they agree in gender, number and case with the nouns they modify. These attributes have almost all the values presented in the relevant sections of Nouns, with the exception of masc-fem and common for Gender, invariant for Number and voc and indcl for Case. Certain forms are monographs, necessitating the use of a disambiguation procedure during corpus tagging:

tou vivliou - neut. gen. tou anthrwpou - masc. gen.

but to exwfullo - neut. nom./acc.

Recommendations 220 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R

Attribute value Gr. example Gr. tag

Gender masculine o AtDfMa

feminine y AtDfFe neuter to AtDfNe

Attribute value Gr. example Gr. tag

Number singular to AtDfNeSg

plural ta AtDfNePl

Attribute value Gr. example Gr. tag

Case nom o AtDfMaSgNm

gen tou AtDfMaSgGe acc to(n) AtDfMaSgAc acc tw AtDfMaSgDt

Recommendations 221 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R 14.10 Application to Swedish Language specific property In Swedish we can express definiteness with enclitic articles. These suffixes, in singular -(e)n, -(e)t and plural -na, will be annotated as a feature of the noun.

14.10.1 Type

Attribute value Example Tag Type indefinite en art-com-sg-indef

indefinite ett art-neut-sg-indef definite den art-com-sg-def definite det art-neut-sg-def definite de art-pl-def

14.10.2 Gender

Attribute value Example Tag Gender nonneuter (en) stol sb-com

neuter (ett) bord sb-neut

Recommendations 222 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R 14.11 Application to Catalan The code AR identifies both definite (el, l', la, els, les) and undefinite (un, una, uns, unes) articles in the lexicon and in the corpus. The article l' has no gender mark, and it does not take any gender tag in the corpus and lexicon.

Recommendations 223 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R 14.12 Application to Irish The article in Irish has a definite form only, an (the), e.g. An Life, The (river) Liffey. (The Irish form is the basis for Joyce's personification Anna Livia). It becomes na in the plural, and in the genitive singular feminine. The masculine and feminine forms are distinguished only by the latter change, and by initial mutations on the noun that follows.

==================================================

Singular Plural Masc. Fem. Masc./Fem. ================================================== Nominative an an+L na Genitive an+L na na+E ==================================================

Recommendations 224 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R 15 Adverb

ADVERB Type Degree Polarity wh-Type

MULTILEX (14 sem true

val.s) false time comp place sup manner ext-c GENELEX quant ext-s

degree compar superl AlethDic

posit NERC compar

superl general posit Q-Type inter Leech degree compar other relat

particle* superl

EAG-L0 ADVERB

general pos EAG-L1 particle comp

sup EAG-L2a ext-c wh

ext-s no-wh En degr En/Sp int Fr ne En/Sp rel Fr pas EAG-L2b Du conj

Ge/Du pron Du sep1 Du sep2

Recommendations 225 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R 15.1 Comments 15.1.1 Type The MULTILEX standard proposes among the "syntactic-properties of LU's" the attribute `Advsubclass', which corresponds to Type, with 13 values based on distinctions of semantic type, plus the value `other' for adverbs other than those indicated by the 13 possible values. For each value application tests and examples are provided (MULTILEX 1993, p.69). In the GENELEX model, the values for Type are proposed at the level of syntactic specifications. The TEI also proposes values on semantic grounds. The feature Type was not proposed in the nucleus of minimal standard specifications by NERC, since most of the analysed tagsets do not make such a distinction. It appears difficult to map existing corpus tagsets onto the subclassifications of adverbs proposed by the values of this feature. Another interesting fact is the inclusion of Particles among the values of the feature Type by the Leech/Wilson proposal. The treatment of Particles appears to be somewhat complicated: NERC suggests their inclusion in the preposition category; in GENELEX and MULTILEX they form a separate category without any attribute.

In the present EAGLES proposal, the values proposed for the feature Type at recommended level reflect a first distinction between general adverbs and particles. All further languagespecific or application-dependent distinctions can be made at level 2b.

15.1.2 Degree The values of this feature are `comparative', `superlative', `positive'. `Positive' can be seen as the default. GENELEX has three more distinctions for each of them.

15.1.3 Polarity This is a language-dependent feature: most English tagsets distinguish between interrogative/relative adverbs (wh-adverbs) and the others. This distinction is also relevant for the Spanish EUROTRA Lexicon.

15.1.4 Other Distinctions We must observe that, even though many richer classifications for Adverbs exist and are used both in lexicons and in tagging practices, these are based on semantic grounds and there is too much disagreement between them. They are, therefore, left as a matter for individual systems.

Recommendations 226 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R 15.2 Application to Italian Adverbs are very similar to the adjective in that, in general, they behave similarly wrt the verb as the adjective does wrt the noun.

Una parola dolce A/ns vs. Parlare dolcemente B/ In traditional Italian grammar, adverbs are subdivided into different subcategories on the basis of their semantic value: manner, time, place, quantity, comparative, affirmative, negative, relative/interrogative, etc.

In corpus tagging and dictionary encoding not all these types are distinguished. 15.2.1 Degree

Attribute value Example Tag

Degree positive bene B

comparative meglio BC superlative benissimo BS

Recommendations 227 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R 15.3 Application to German 15.3.1 Type There are three types of adverbs:

ffl ADV: general (i.e. non-adjectival , cf. section ??) ffl PWAV: interogative or relative ffl PROAV: pronominal adverbs. These are da or hier + preposition (such as damit, daf"ur,

dagegen, daneben etc), which replace a PP with the specific preposition.

Example: er kommt heute/ADV

er kommt schnell/ADJNA wo/PWAV wohnt er? der Ort, wo/PWA er wohnt er besteht darauf/PROAV

Semantic distinctions like time, place, etc. are not expressed in the annotations.

Attribute value example tag

Type general oft ADV

int/rel wo PWAV

pronominal hierbei PROAV

15.3.2 Degree of comparison (This feature is not used for adverbs in the IMS-Tagset as it applies only to a very few nonadjectival adverbs.)

Attribute value example tag

Degree positive oft ADV:Pos

comparative "ofter ADV:Comp superlative (am) "oftesten ADV:Sup

Recommendations 228 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R 15.4 Application to English In English, adverb particles are an important category because of their role in the forming of phrasal verbs (compare separable verb forms in German and Dutch: 3.5.11).

15.4.1 Adverb-type

Attribute values Examples Tags Adverb-type general quickly AV

degree very AVD particle up AVP

15.4.2 Degree

Attribute values Examples Tags Degree positive soon AV

comparative sooner AVR superlative soonest AVT

15.4.3 Polarity

Attribute values Examples Tags Polarity wh-type where? AVWQ

other here AV

15.4.4 Wh-subtype

Attribute values Examples Tags Wh-subtype relative when AVWR

other when? AVWQ

(Compare Pronoun and Determiner above.) (Similar distinctions are made under Pronoun and Determiner above, 5.4.2 and 6.3.1).

Recommendations 229 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R 15.5 Application to Dutch 15.5.1 Type: Language-specific features Dutch grammar distinguishes two subcategories of adverbs. One category is called pronominal adverb, because it has the same referential function as a pronoun. The other is called conjunctival adverb, because it functions as a conjunction.

The table below is a proposal for Dutch, not a CELEX table !

Attribute value Example Tag

Type normal (hij kmot) vaak Adv

pronominal (Ik reken) erop PronAdv

daar (houk ik van) SepPronAdv conjunctival bovendiven (is hij ziek) ConjAdv

waar (het betref)

As pronominal adverbs are separable word forms, Dutch needs a way of marking up the following three possibilities:

The pronominal adverb as a whole: Tag = PronAdv The first part of a separable pronominal adverb: Tag = SepPronAdv The second part of a separable pronominal adverb: Tag = SepAdp (see Adpositions).

This can be an L2b tag.

15.5.2 Degree of comparison CELEX does not have this distinction at adverbial level. It does have this information at adjectival level.

Recommendations 230 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R 15.6 Application to Spanish 15.6.1 Type (semantic)

Attribute value Example Tag

Type general

degree

The Spanish Eurotra dictionaries include "semantic typing" at the syntactic level in order to deal with several phenomena related to the attachment of modifiers. The feature used and its values are as follows:

vadv = time, manner, deg(ree), loc(ation), free, manconcret. Also, in relation to overgeneration in attachment, some features control the attachment of adverbs in order to prevent sentences as:

* Todo termin'o desgraciadamente (Everything finished unfortunately)

Desgraciadamente aquello era verdad (Unfortunately that was true)

15.6.2 Degree

Attribute value Example Tag

Degree positive muy

comparative mejor superlative

In the ET-ES dictionaries there is no attribute to code the values of "degree". In Spanish, the comparative of adverbs is formed analytically with tan, m'as, or menos: m'as r'apido que yo (faster than I) m'as pronto (soonner)

Only four adverbs are have irregular comparative morphology: bien/mejor (well/better) mal/peor (badly/worse) mucho/m'as (much/more) poco/menos(little/less)

Recommendations 231 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R 15.6.3 Polarity The Eurotra dictionaries have the attribute "whmor", with the three possible values "int", "rel" and "none", which serves to distinguish between interrogative, relative and other adverbs:

Attribute value Example Tag

Polarity muy Wh-Type int d'onde, cu'ando, c'omo whmore=int

rel donde, cuando, whmore=rel non-Wh aqu'i, ahora, as'i whmore=none

15.6.4 Apocope Some adverbs (just as we saw wrt adjectives) change their form when they occurr before an adjective or another adverb:

tanto/tan (ricamente) cuanto/cuan (largo) mucho/muy (despacio)

Recommendations 232 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R 15.7 Application to French (Corpus) As in other languages, French has a tradition of distinguishing classes of adverbs (place, manner, etc.) on the basis of semantic considerations. Degree is one of these classes. However, the IBMF tagset has a unique tag for adverbs, ADVE (with the exception of negative particles, see below).

15.7.1 EAGLES features not applicable Polarity and wh-type do not apply to French.

15.7.2 IBMF Tagset features not applicable in EAGLES The tagset has two tags for specific negation adverbs: NE and PAS (the latter class containing pas and plus).

15.8 Application to French (Lexicon) 15.8.1 Type ------------ ----------- ----------- ---- Attribute Value Example Code ------------ ----------- ----------- ---- Type general fortement g

particle ne, pas p ------------ ----------- ----------- ----

It seems necessary in French to distinguish the two parts of the negation ("ne ... pas"), because they play an important role in disambiguation.

15.8.2 Degree ------------ ----------- ----------- ---- Attribute Value Example Code ------------ ----------- ----------- ---- Degree positive fortement p

comparative davantage c ------------ ----------- ----------- ----

We encode Degree for compatibility with other languages, but, as with adjectives, the comparative feature is not very productive in French. It applies only to beaucoup (comp.= "davantage"), bien (comp.= "mieux"), "mal" (comp.= "pis") and "peu" (comp. = "moins"). The comparative for other adverbs is marked by "plus" + adverb (e.g. "plus fortement"). The superlative is usually marked by "le"+ comparative (e.g. "le plus fortement").

15.8.3 Combinations --------- ----------- Tag Example --------- ----------- Rgp beaucoup

Recommendations 233 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R Rgc davantage Rpn ne Rpn pas, plus --------- -----------

Recommendations 234 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R 15.9 Application to Portuguese 15.9.1 Type Not used in the Portuguese morphological application.

15.9.2 Degree ------------ --------------- ----------- ---- Attribute Value Example Code ------------ --------------- ----------- ---- Degree positive

comparative superlative ============ =============== =========== ==== l-specif comparative+ mais

comparative- menos comparative= ta~o ------------ --------------- ----------- ----

In the Portuguese model the `positive' value is not explicitly marked because it is considered a default value.

Recommendations 235 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R 15.10 Application to Danish In traditional Danish grammars the word class of adverbs is divided into subclasses according to their semantic content and possible position within a sentence (field grammar). Dictionaries do not deal with semantic subclasses. The EDEMD distinguishes between real adverbs and derived adverbs. Furthermore, distinctions are made on the basis of the syntactic properties of adverbs (modifying role, scope and position in sentence). Generally, in corpus tagging and dictionary encoding the above mentioned distinctions are not used. For real adverbs (i.e. not derived from adjectives) there are only a few relics of a system of degree comparison. Most grammars regard adverbs as uninflectable.

15.10.1 Degree

Attribute value Example Tag

Degree positive ofte adv pos

comparative oftere adv comp superlative oftest adv sup

Recommendations 236 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R 15.11 Application to Greek 15.11.1 Type

Attribute value Gr.Example Gr.Tag

Type

l-spec sent isws AvSe l-spec nonsent argha AvNs l-spec both kala AvBt

Traditionally, in Greek, adverb types correspond to a semantic subcategorisation. Thus, the values they usually have are: manner, time, place, quantity, comparative, negative, affirmative, etc.

However, for computational applications, we have considered more appropriate to distinguish between two different categories, on the basis of scope: sentential adverbs, which have scope over the whole sentence, and nonsentential ones, which usually modify a verb, adverb or adjective.

Certain adverbs can be used both sententially and non-sententially, in which case they are coded with the value both in the Lexicon; disambiguation during corpus processing relies on the linguistic context (most usually, the presence or absence of commas):

O Ghiannys milaei kala. - nonsent.

Kala, autos twra ti thelei; - sent.

15.11.2 Degree As in the case of adjectives, not all adverbs in Greek form the comparative and superlative degree. In the same way, they form these two degrees either monolectically or periphrastically. (See also relevant section on adjectives.)

Attribute value Gr.Example Gr.Tag

Degree basic akriva AvNsBa

comparative akrivotera AvNsCp superlative akrivotata AvNsSu

15.11.3 Function This is a language-specific feature, which should be included at level L2a of the EAGLES proposal. It has been introduced in order to generalise over certain characteristics that Greek adverbs present, thus allowing them to be grouped together and, in this way, avoid increase of ambiguity during corpus tagging and prepare the grounds for syntactic analysis.

Recommendations 237 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R More specifically, we distinguish between five types of adverbs: (a) Pure adverbs : they behave only as adverbs, i.e. modifying verbs, adjectives, sentences, or other adverbs.

(b) Conjunctive adverbs : under this category, we include anaphoric adverbs, which are traditionally considered as adverbs but which, unlike pure adverbs, also introduce a secondary sentence:

Kathws anaferei o sughghrafeas, o sevasmos einai aghrafos nomos. (c) Interrogative adverbs : they behave as conjunctions with the further constraint that they can only introduce main interrogative or exclamatory sentences or secondary interrogative clauses:

Ghiati efughes? - main int. Ti wraia pou ta perasame! - main excl.

Dhen xerw ghiati efughe. - sec.int.

(d) Comparative adverbs : they introduce comparison and, unlike pure adverbs, they may be followed by a phrase of the same cateogory as the standard of the comparison, a noun/adjectival phrase which has the same case as the standard of the comparison, a secondary sentence introduced by the particle "na" or (as regards the adverb "san" only) a phrase in the accusative case:

Etrwghe san vasilias. - noun in nominative Etrwghe san to vasilia. - noun in accusative Etrwghe sa na ytan vasilias. - secondary clause

(e) Prepositional adverbs : this class includes mainly local, and a few manner and time adverbs that can sometimes be followed by a clitic or a noun phrase in the genitive case :

Estripse aristera. - nothing Aristera sou vrisketai ena kouti. - clitic Dhiavazei to "Anatolika tys Edhem". - noun phr. in gen.

Attribute value Gr.Example Gr.Tag

Function pure akriva AvNsBaPu

conjunctive opws AvNsBaCj interrogative ghiati AvNsBaIr comparative opws AvNsBaCo prepositional aristera AvNsBaPp

Note that, as shown in the above table, certain adverbs may belong to more than one classes. For instance, "opws" is coded as both a conjunctive and a comparative adverb. Disambiguation can only be performed on the basis of the semantic function of this word in the specific linguistic context it occurs:

Recommendations 238 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R

Opws sas einai ghnwsto, efughe. - conj. Auto ishuei ghia periptwseis opws y aupnia. - comp.

15.11.4 Features not applicable in Greek Polarity and wh-Type are not applicable to Greek adverbs.

Recommendations 239 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R 15.12 Application to Swedish 15.12.1 Degree

Attribute value Example Tag Degree positive ofta adv-pos

comperative oftare adv-comp superlative oftast adv-sup

Recommendations 240 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R 15.13 Application to Catalan The code AV identifies adverbs. There are not subtyped features. Catalan adverbs can take appreciative suffixes, such as aviat-aviadet.

Recommendations 241 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R 15.14 Application to Irish The adverbial system reproduces most features of the adjectival system and the base form of the adverb is formed simply by putting the particle go before the adjective.

15.14.1 Type ======================================================================== Attribute Value Example Tag ======================================================================== Type General go maith, go dona [well, poorly] AV

Degree an-, ro', sa'ch [very, too, moderately] AVD Phrasal (dul) amach, isteach, suas [(to go) out, in, up] AVP Wh-form conas, cathain, ca'r [how?, when?, where?] AVW Rel-form mar, nuair (where, when) AVR ========================================================================

15.14.2 Degree ======================================================================== Attribute Value Example Tag ======================================================================== Degree Base (labhair) go hard [(speak) loudly] AV

Compar (labhair) ni'os airde [(speak) more loudly AVC ========================================================================

Recommendations 242 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R 16 Adposition

ADPOSITION Type Formation Gend Numb Case

MULTILEX preposition

postposition circumposition GENELEX Preposition*

AlethDic Preposition*

NERC Preposition*

Leech preposition

postposition

EAG-L0 ADPOSITION

preposition simple m sg EAG-L1 postposition fused f pl

circumposition n

nom EAG-L2a gen

dat acc EAG-L2b Du 2nd-p It c

Du sepadv

Recommendations 243 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R 16.1 Comments 16.1.1 Type The values of this feature allow one to distinguish between `preposition' and `postposition'. MULTILEX also has a value for circumposition (see MULTILEX). In NERC, GENELEX and AlethDic, Prepositions constitute a separate category without any attribute at the morphosyntactic level; in GENELEX the distinction based on a semantic classification between `temporal', `manner' and `place' is foreseen at the syntactic level.

In NERC, Prepositions also include Particles. In English there is a problem of distinguishing the verb particles from prepositions, but many tagsets provide a separate label for particles.

16.1.2 Formation In some languages, e.g. in Italian (see below), prepositions may appear fused with articles in a unique graphical form. EAGLES encourages the treatment of these forms as separable, therefore having two tags corresponding to the two different categories, but also allows the use of a unique tag.

This general recommendation, i.e. to separate different categories and to encode them with different tags, is also applicable to other cases of contractions, such as negation, cliticisation, etc.

In general, there will be language-specific options for the treatment of these phenomena.

Recommendations 244 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R 16.2 Application to Italian 16.2.1 Type

Attribute value example It.tag Type preposition di, a, da E

In Italian, simple prepositons are distinguished in two categories: (i) a closed set, di, a, da, in, con, su, per, fra, tra which are used to express a number of syntactic relations. (ii) a larger set with more specific meanings, e.g. sopra, sotto, prima, dopo, senza, etc.. These constitute a crux in tagging because they are involved in transcategorization phenomena (see Conjunctions). The tag is E.

Most of the prepositions in category (ii) require the presence of another preposition before the argument, sopra di noi: this raises the problem of which strategy to choose in multiword expression tagging (see Leech and Wilson Invitation Draft).

16.2.2 Formation Prepositions in category (i) are often fused with following articles, forming the so-called preposizioni articolate: ex. agli (a+gli), ai (a+i), degli (di+gli), dello (di+lo). These compounds, which can be found instantiated only in a corpus, are tagged with the tag (E) plus the morphological features of the fused article.

Attribute value example It.tag Formation simple di, a, da E

fused dagli, agli E/mp

Recommendations 245 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R 16.3 Application to German Three types of adpositions are distinguished: preposition, postposition and circumposition. There are a few cases where prepositions and a following definite article are contracted:

Example: zum = zu + dem

ans = zu + das

16.3.1 Type

Attribute value example tag

Type preposition ohne (mich) APPR:Akk

postposition (ihm) zuliebe APPO:Dat circumposition von (dem Tag) an APZR8

16.3.2 Formation

Attribute value example tag Formation simple in (das Haus) APPR:Akk

zu (den Leuten) APPR:Akk fused ins (Haus) APPRART:Neut.Akk.Sg

zur (Schule) APPRART:Fem.Dat.Sg

8Only the second part of the circumposition is actually annotated with APZR: von/APPR:Dat dem Tag an/APZR.

Recommendations 246 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R 16.4 Application to English 16.4.1 Adposition-type

Attribute values Examples Tags Adposition-type preposition at APR

postposition 's APO

On the classification of the genitive particle 's as a postposition, see Nouns above.

Recommendations 247 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R 16.5 Application to Dutch 16.5.1 Type CELEX has no Adpositions of type preposition. It only has straightforward Prepositions as a part of speech. However the German table with attribute values: preposition, postposition and circumposition is applicable to Dutch as well. But, as mentioned above, we need a special language-specific attribute value to mark up the non-verbal parts of separable verb forms and another to mark up the second part of separable pronominal adverbs: The table below is a proposal for Dutch, not a CELEX table !

Attribute value Example Tag

Type preposition zonder (mij) PreP

postposition (de berg) op PostP circumposition van (die dag) af CircumP 2nd part of pron.adv. (daar houd ik) van SepAdp non-verb. part of sep.adv. (gaven) aan SepVAdp

Recommendations 248 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R 16.6 Application to Spanish 16.6.1 Prepositions In our dictionaries there are no special attributes for prepositions at the morphological level. At the configurational level, a semantic typing of prepositions helps to control attachment of modifiers.

ptype: norm, loc, orig, dest, nil In Spanish some contractions of articles and prepositions can be found: a + el = al de + el = del

For analysis purposes, these contractions are treated at the morphological level in ET-ES grammars as samples of a special category "portmanteau", and are decomposed before reaching the configurational level ECS.

Recommendations 249 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R 16.7 Application to French (Corpus)

Attribute value Example Tag Type preposition dans PREP

Attribute value Example Tag Formation simple dans PREP

fused au PAU

The IBMF tagset has 2 classes of preposition: PDEA for `a and de

PREP for other prepositions

and 4 classes of preposition-determiner: PAU for au (`a + le) PAUX for aux (`a + les)

PDES for des (de + les) PREPMS for du (de + le)

16.8 Application to French (Lexicon) 16.8.1 Type ------------ ----------- ----------- ---- Attribute Value Example Code ------------ ----------- ----------- ---- Type preposition en, de p ------------ ----------- ----------- ----

16.8.2 Combinations --------- ----------- Tag Example --------- ----------- Sp en --------- -----------

Recommendations 250 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R 16.9 Application to Portuguese 16.9.1 Type ------------ ---------------- ----------- ---- Attribute Value Example Code ------------ ---------------- ----------- ---- Type preposition em

postposition circumposition ------------ ---------------- ----------- ----

16.9.2 Formation This feature is not used in the Portuguese lexicon, but it is pertinent to the Portuguese language, where contraction phenomena occur. Following the GENELEX multilingual model, in the Portuguese lexicon contraction phenomena occuring typically with preposition plus determiner are dealt with as a special kind of morphological unity (the unit'e morphologique agglutine'e.

Recommendations 251 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R 16.10 Application to Danish In automatic corpus tagging the recognition and annotation of circumpositions is a special task to be dealt with. At present there does not exist a commonly agreed tagset for the three types of adpositions: simple preposition e.g. til, complex preposition e.g. uden for and circumposition e.g. for ... skyld. Furthermore, as in English, the genitive 's may be considered as an enclitic postposition. The table below may be regarded as a proposal for Danish.

16.10.1 Type

Attribute value Example Tag

Type preposition, simple til p s

preposition, complex uden for p c circumposition for (denne sags) skyld p cc

Recommendations 252 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R 16.11 Application to Greek 16.11.1 Type

Attribute value Gr. example Gr.tag Type preposition me, kata Pp

As indicated by the above table, prepositions in Greek actually form a category of their own. In Greek we distinguish between two kinds of prepositions: (a) simple ones, included in this category, and (b) complex ones, formed by the combination of an adverb and a preposition: e.g. panw apo, mazi me, etc.. These can only be recognised at text level with the use of a tagger that looks above word level.

The category of simple prepositions constitutes a closed set, while the latter is productive. At present, we make no effort at recognising complex prepositions, given the need of a multi-word tagger, although such a mechanism is planned for the near future.

16.11.2 Form, Gender, Number and Case One of the simple prepositions, namely se, when followed by the definite article, it is fused with it and forms what we call prepart: ston (se + ton), styn (se + tyn), sto (se + to), stous (se + tous), stis (se + tis), sta (se + ta). These forms are further coded for the morphological features of the article (i.e. Gender, Number and Case). Note that for these features, not all case values are possible (only the genitive and accusative).

Attribute value example Gr.tag Form simple se PpSp

prepart ston PpPa

Attribute value Gr. example Gr. tag

Gender masculine ston PpPaMa

feminine styn PpPaFe neuter sto PpPaNe

Recommendations 253 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R

Attribute value Gr. example Gr. tag

Number singular ston PpPaMaSg

plural stous PpPaMaPl

Attribute value Gr. example Gr. tag

gen stou PpPaMaSgGe acc sto PpPaMaSgAc

16.11.3 Function A number of prepositions in Greek display a deviation from the typical syntactic behaviour of prepositions. As "typical" behaviour we consider the cases where the preposition is followed by :

(a) a nominal phrase in an oblique case (i.e. mainly the accusative case, more rarely, the genitive, and, only for a few ancient prepositions that are still but rarely used in the Modern Greek language, the dative case):

Vghykame me to Ghianny. - acc. Ftasame ekei dhia tys plaghias odhou. - gen. En suneheia, tha suzytysoume ektenestera to thema auto. - dat.

(b) an adverb - only for the prepositions "ghia", "apo" and "kata":

Yrthame apo ekei. (c) a clitic pronoun - only for the prepositions "metaxu", "anametaxu", "uper", "enantion", "enwpion":

Ta dhylwse enwpion mas. On the borderline of this behaviour, we find two groups of elements: 1) A group of six prepositions,namely "ghia", "mehri", "hwris", "dhihws", "isame" and "anti", can be followed by the particle "na" and, in this case, they show similar behaviour to that of conjunctions given that they introduce secondary sentences:

Tha kanei ta panta ghia na fughei o Kwstas.

Efughe hwris na mas hairetysei.

Recommendations 254 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R In order to account for the particularity of this group, the feature function has been introduced in the ILSP tagset. Therefore, the distinction holds between pure and conjunctive prepositions. This feature is intended to be used at a later stage, for syntactic processing.

Attribute value Gr. example Gr. tag

Function pure me PpSp000000Pu

conjunctive ghia PpSp000000Cj

2) A group of elements which, in addition to the "typical" for prepositions linguistic contexts, may also be found in syntactic contexts considered typical of adverbs, i.e.:

(a) standing on their own, or (b) followed by a preposition:

Mpyke meta to gheuma. - noun phrase in acc.

Mpyke meta. - nothing Mpyke meta apo to gheuma. - prep. phrase

We decided to use the syntactic behaviour as a criterion for distinguishing between prepositions and adverbs, given that in Greek these two are easily confused, granted that prepositions slide over time from the category of prepositions to that of adverbs. Thus, those elements that display both behaviours are coded twice in the Morphological Lexicon, as adverbs and prepositions, and disambiguation during corpus tagging is performed on the basis of a rule taking advantage of the linguistic context:

Mpyke meta to gheuma. - prep. Mpyke meta apo to gheuma. - adv.

Mpyke meta. - adv.

There is one exception to this rule, namely "anti", which is considered always a preposition, although it can be followed by a preposition. This preposition is considered a special case for two reasons: (a) it can never stand alone, in the way that adverbs do, (b) it is the whole compound "anti ghia" that acts as a preposition, having the same meaning as the simple "anti", (c) when followed by any other preposition besides "ghia", it is due to elliptical speech, as it is the same preposition that introduced the element that is replaced:

Anti tou Ghianny yrthe y Maria. - gen. Anti ghia to Ghianny yrthe y Maria. - prep. "ghia" Aghorase apo to mpakaly anti apo to manavy. - prep.

Recommendations 255 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R 16.12 Application to Swedish 16.12.1 Type

Attribute value Example Tag Type preposition till p-pre

circumposition f"or (den sakens) skull p-cc postposition (v"anner) emellan p-post

Recommendations 256 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R 16.13 Application to Catalan The code PO identifies prepositions. There are not subtyped features.

Recommendations 257 May, 1996 EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R 16.14 Application to Irish Irish has prepositions only. Language specific features are the attributes Person, Number and Gender.

16.14.1 Person ======================================================================== Attribute Value Example Tag ======================================================================== Person Null do [to] P

First dom, du'inn [to me/us] P1s/p Second duit, daoibh [to you (sing/plural)] P2s/p Third do', di, do'ibh [to him/her/them] P3sm/f, P3p ========================================================================

16.14.2 Number ======================================================================== Attribute Value Example Tag ======================================================================== Number Null ar [on] P

Sing orm, ort, air, uirthi [on me/you/him/her] P1/2/3sm/f Plural orainn, oraibh, orthu [on us/you (pl.)/them] P1/2/3p ========================================================================

16.14.3 Gender ======================================================================== Attribute Value Example Tag ======================================================================== Gender Null o' [from] P

Masc uaidh[from him] P3sm Femin uaithi [from her] P3sf ========================================================================

Recommendations 258 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R 17 Conjunction

CONJUNCTION Type Coord-T Subord-T Iter Sent.Intr

MULTILEX Coord* simple yes yes

Subord* repetit no no

correlat GENELEX coord

subord AlethDic coord

subord NERC coord

subord Leech coord simple

subord repetit

correlat sent.init

EAG-L0 CONJUNCTION EAG-L1 coord

subord EAG-L2a +infve

compar +fin EAG-L2b En simple

En init En no-init Sp correl

Recommendations 259 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R 17.1 Comments Conjunctions link syntactically two or more words or two or more syntagms: phrases, sentences, etc.

Coordinating: determines a syntactic equivalence between conjuncts:

Mary and John Subordinating: links two sentences in a dependency relation:

I do not answer, since I was ... A large core of agreement emerges as regards Conjunctions. All the analysed systems agree as to the distinction between coordinating and subordinating Conjunctions. The only difference is the MULTILEX position which splits the two types of Conjunctions into two different categories, `coordinators' and `subordinators': however, no mapping problems arise. The same choice was made by the TEI.

17.1.1 Coordination Type Within the class of coordinators MULTILEX foresees some features pertaining to the syntactic level, which in the Leech/Wilson proposal are collected under a unique feature Coord-Type: - `simple': between conjuncts, (John and Mery), - `repetitive': before each conjunct, (both John and Mary), - `correlative': before a conjoined phrase, it requires specific coordinators between conjuncts, (either John or Mary). `Repetitive' and `correlative' imply the specification of Expected Coordinator, i.e. both requires and etc., and the specification of Iteration, i.e. whether a coordinator can coordinate more than two conjuncts, and ... and ... and (more than two: yes), but (two: no). These last two attributes are introduced by MULTILEX only. Sentence Introducer is used for coordinators which can introduce a sentence, linking it by coordination to the preceding sentence: And Mary left. (Either cannot introduce a sentence in this way).

In the EAGLES Proposal these further distinctions concerning the Type of Coordination are treated as language-specific features.

17.1.2 Subordination Type Further information on the Type of Subordination, i.e. if the conjunction requires a finite verb "+fin", a non-finite verb "+infve" or introduces a comparison "compar" are proposed on the Level-2b, since they are shared by three different languages.

Recommendations 260 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R 17.2 Application to Italian Conjunctions in Italian constitute an "open class": the same elements can often have the function of conjunctions, prepositions, adverbs.

Conjunction: Ti sentirai tranquillo, dopo aver sostenuto l'esame Preposition: Ti sentirai tranquillo dopo l'esame Adverb: Sostenuto l'esame, dopo ti sentirai tranquillo

For these classes, as far as possible, application tests have to be employed: - preposition: the argument is an NP; - conjunction: the sentence introduced by the elements is in a dependency relation with the governing predicate; - adverb: it semantically modifies the whole sentence and the predicate in particular.

In grammars, no agreement is reached as to the inclusion of some elements in the conjunction or the adverb class (e.g. anche, pure, dunque, pertanto).

In some recent linguistic theories, all these elements go into the wider class of "connettivi" (connectives), which link various parts of a text (Berretta 1984). In tagging Italian the problem of subordinating "locutions" (multi-words) has to be dealt with, e.g. dal momento che.

17.2.1 Type

Type value example It.tag

subordinating perche' C coordinating e CC

The dictionary proposes the distinction between the coordinating and subordinating function:

Maria e Giovanni CC non ho risposto perche' ero disattento C

Recommendations 261 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R 17.3 Application to German 17.3.1 Type

Attribute value example tag

Type coordinating (ich) und (du) KON

subordinating (er will,) dass KOUS comparison (kleiner) als KOKOM

17.3.2 Subord-Type

Attribute value example tag Subord-Type finite (er kommt,) wenn KOUS

with infinitive ohne (zu wollen) KOUI

Recommendations 262 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R 17.4 Application to English 17.4.1 Conjunction-type

Attribute values Examples Tags Conjunction-type coordinating and, or CC

subordinating if CSF

17.4.2 Coord.-type

Attribute values Examples Tags Coord.-type simple and, or CC

initial both (...and) CCI non-initial (neither...) nor CCM

17.4.3 Subord.-type

Attribute values Examples Tags Subord-type with finite verb because CSF

with non-finite verb in order (to) CSN comparative than CSC

Recommendations 263 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R 17.5 Application to Dutch The German scheme is applicable to Dutch as well, but there is no such application in CELEX.

Attribute value Example Tag

Type coordinat (ik) en (jij) CoorConj

postposition (hij wil) dat SubConj subord + inf.ve zonder (te willen) InfConj comparison (kleiner) dan CompConj

Recommendations 264 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R 17.6 Application to Spanish In Eurotra coordinating and subordinating conjunctions are treated as two different categories. While at the morphological level they have no special attributes, at the constituent level we have three attributes for the subordinating type:

- sconjtype (valued as circ, compar, compl), which serves to distinguish between: circumstantial, e.g.: MIENTRAS yo trabajo, el juega (while I work, he plays) comparative, e.g.: Juan es tan lento COMO su hermano (John is as slow as his brother) complementizers, e.g.: Juan dice QUE es verdad (John says that it is true)

- ftranc, to lexically encode the value of the conjunction to make up a transconstructional modifier. 9 - comp (valued as q, nonq). Complementizers are thus distinguished with respect to their interrogative value: "que/si" (that/whether)

17.6.1 Type

Attribute value Example Tag

Type coord Juan Y Maria cconj

subord no se SI vendr'e sconj

cconj: y, pero, ni, o sconj: cuando, si,que, porque, aunque, mientras, etc.

17.6.2 Coord-Type

Attribute value Example Tag

Coord-T simple

correlat NI Juan NI Mar'ia initial non-initial

Coord-T distinctions are not coded in the ET-ES dictionaries. There are no examples for "non-initial" coordinating conjunctions.

9Definition of Transconstructionals: The Eurotra Reference Manual 7 defines transconstructionals as those constituents which do not modify the governor of the construction to which they belong, but provide information about the sentence as a whole.

Recommendations 265 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R 17.6.3 Subord-Type

Attribute value Example Tag Subord-T +infve para

+fin mientras comparison tan

In the ET dictionaries, there exists the feature "sform" with values infin, finite, pastpart and gerund, which is also used to characterise subordinating conjunctions. For those subordinating conjunctions which lexically require finite verbal forms, Spanish also requires information on the mood of the subordinate verb. To that end, the ET-ES lexica encode a feature "exig-mood", valued "indicative/subjunctive", mainly for synthesis purposes.

Recommendations 266 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R 17.7 Application to French (Corpus)

Attribute value Example Tag Type coord et CCOO

subord que CSUB

All EAGLES attributes apply in principle to French. Only the coordination/subordination distinction is coded in the tagset.

17.8 Application to French (Lexicon) 17.8.1 Type ------------ ----------- ----------- ---- Attribute Value Example Code ------------ ----------- ----------- ---- Type coordinat. et c

subordinat. que s ------------ ----------- ----------- ----

17.8.2 Combinations --------- ----------- Tag Example --------- ----------- Cc mais, ou Cs que --------- -----------

Recommendations 267 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R 17.9 Application to Portuguese 17.9.1 Type ------------ ------------- ----------- ---- Attribute Value Example Tag ------------ ------------- ----------- ---- Type coordination e

subordination que ------------ ------------- ----------- ----

Recommendations 268 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R 17.10 Application to Danish In general, the EAGLES attributes apply also to Danish, thus the scheme for English is applicable to Danish as well. Traditional dictionaries do not distinguish between conjunction types, but grammars do. The EDEMD uses the following additional features, as compared with EAGLES, within the description of coordinating conjunctions: conjunction (simple coordination) e.g.og, disjunction e.g.eller and adversative conjunction e.g.men. Furthermore, a negation inherent to the conjunction is also catered for in the encoding guidelines. At present, there is no common definition of appropriate tags for corpus annotation purposes.

17.10.1 Type

Attribute value Example Tag Conj.Type coordinating og coconj

subordinating fordi subconj

Attribute value Example Tag Coord.Type simple og coconj

initial hverken (... eller) coconj init non-initial (hverken ...) eller coconj n-init

Attribute value Example Tag Subord. Type with finite verb fordi subconj vfin

with infinitive for at subconj inf comparative end subconj comp

Recommendations 269 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R 17.11 Application to Greek Under this category, we have included only "pure" conjunctions, i.e. elements that are used to introduce main or secondary sentences, and have avoided the inclusion of prepositions and adverbs that function as conjunctions as well. These have been treated in the respective categories. (For further details, please refer to the relevant sections.)

Like prepositions, we can distinguish between two kinds of conjunctions in Greek: (a) a finite set of simple conjunctions: e.g. kai, alla, myn, oute, and (b) a productive set of conjunctions formed periphrastically, with some word/phrase combined by a conjunction: e.g. dhedhomenou oti, an kai, etc. The tagging of these multi-word forms must be performed by a special mechanism.

17.11.1 Type

Type value Gr.example Gr.tag

coord kai CjCo subord oti CjSb

Conjunctions in Greek can be further subcategorised depending on their semantic value: causal, temporal, enumerating, etc. However, this distinction was not considered necessary for the present codification.

17.11.2 Coordination Type This feature is not presently coded in the Greek Lexicon, but applies to the language.

We distinguish two cases: - conjunctions that are always found in the text alone: e.g. alla, - conjunctions that can be found alone or combined with another conjunction before the first conjunct: e.g. kai, y.

O Ghiannys kai y Maria efughan.

or Kai o Ghiannys kai y Maria efughan.

In this case, the values simple and correl proposed at level L2b could be used for Greek.

Recommendations 270 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R 17.12 Application to Swedish 17.12.1 Conj.Type

Attribute value Example Tag Conj.Type coordinating och coconj

subordinating eftersom subconj

17.12.2 Coord.Type

Attribute value Example Tag Coord.Type simple och coconj

initial varken (...eller) coconj-init non-initial (varken ...) eller coconj-n-init

Recommendations 271 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R 17.13 Application to Catalan The code C identifies conjunctions. No subtyped features.

Recommendations 272 May, 1996 EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R 17.14 Application to Irish 17.14.1 Type ======================================================================== Attribute Value Example Tag ======================================================================== Type Coord agus, no' [and, or] CONC

Subord mar, ach [since, but] CONS ========================================================================

17.14.2 Language specific features Among the subordinating conjunction belong the complementizers a+L, a+E, and go (that, sometimes meaning who, which) which make up a complex system of subordination that subsumes the function of relative pronouns in other languages, but is used to achieve other forms of subordination also. This explains the absence of a value relative on pronouns above.

Recommendations 273 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R 18 Numeral

NUMERAL Type Gender Number Case Function

MULTILEX Cardin*

GENELEX cardin* m s Det

ordin* f p AlethDic cardin* m s Adj

ordin* f p NERC cardin

ordin cardin m s nom Leech ordin f p gen

denominat n dat frequency c acc

EAG-L0 NUMERAL EAG-L1 cardin m sg

ordin f pl

n EAG-L2a EAG-L2b Du quantif It c It n Gr nom Pron

denomin Gr gen Det

Gr dat Adj Gr acc Gr voc Gr indc

Recommendations 274 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R 18.1 Comments The analysed systems show a diverging treatment of the Numeral category. Not all languages and systems may want to handle them as a separate category. EAGLES leaves open the option of handling them in the respective relevant classes of Pronouns, Determiners and Adjectives.

18.1.1 Type In MULTILEX cardinals form a separate category without any attribute.

GENELEX and AlethDic propose a distinction between cardinals and ordinals within the Adjective category; furthermore, GENELEX has a cardinal value in the `Determinant' category.

18.1.2 Function In EAGLES, Numerals are proposed as a separate category and the attribute Function is also introduced in order to deal with systems which presuppose a further distinction between Pronouns and Pronominal Adjectives (Italian), and Determinant and Adjectives (French, GENELEX).

Recommendations 275 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R 18.2 Application to Italian In Romance language practices (French, Italian, etc.) Numerals constitute a subclass of Pronouns and Pronominal Adjectives.

They are subdivided into cardinal and ordinal: - cardinal indicates absolute numeral quantity: zero, cento, mille; - ordinal indicates order and classification: primo, secondo

18.2.1 Type

Attribute value It. example It. tag

Type cardinal zero N

ordinal secondo PN/ms

18.2.2 Gender

Attribute value It. example It. tag

Gender masculine primo (anno) DN/m

feminine (la) seconda PN/f

The value common does not apply to Italian ordinals, as they behave as adjectives with due uscite (two endings).

18.2.3 Number

Attribute value It. example It. tag

Number singular prima PN/fs

plural prime PN/fp

18.2.4 Function

Attribute value It. example It. tag

Function pronoun (la) prima PN/fs

determiner primo amore DN/ms

The following are examples of both the pronominal and determiner functions in Italian:

non voglio la prima fetta DN/fs, dammi la seconda PN/fs

Recommendations 276 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R 18.3 Application to German In the IMS-Tagset only cardinals are distinguished from other POS-Types. Ordinals are classified as adjectives.

Example: der dritte/ADJA Mann

drei/CARD M"anner im Jahre 2000/CARD

18.3.1 Numerals: Type

Attribute value example tag

Type cardinal 27 CARD

ordinal - -

Recommendations 277 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R 18.4 Application to English Gender and Case do not apply to English numerals.

18.4.1 Numeral-type

Attribute values Examples Tags Numeral-type cardinal two, 55 NUC

ordinal second, 55th NUO

18.4.2 Number

Attribute values Examples Tags Number singular one NUCs

plural fifty NUCp

Note that when a numeral word occurs with a plural inflection (firsts, fifties), this is considered a clear indicator of the word's status as a noun, rather than as a numeral in the normal sense. All numerals can be converted into nouns for special functions.

Recommendations 278 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R 18.5 Application to Dutch CELEX has a morphological word class tag Q, used in derivational word analysis, which covers Numerals and Quantifiers. Apart of that, CELEX has a syntactical word class tag NUM (Numerals) divided in two subclasses: cardinals and ordinals.

18.5.1 Type

Attribute value Example Tag

Type Quantifier veel NUM quant

Cardinal zeventien NUM hoofd Ordinal zeventiende NUM rang

This is not a CELEX tag. We decided to include it in the table, though, because Dutch traditional grammarians and lexicographers distinguish a subclass, called `onbepaalde (hoofd)telwoorden' in Dutch. Some call them `indefinite numerals', others `indefinite cardinals'. This class concerns words like English `many', `more' , `most' and the like.

Recommendations 279 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R 18.6 Application to French (Corpus) The IBMF tagset has only one tag for numerals, CHIF. However, most of the attributes distinguished in EAGLES apply to French, except case, as usual.

18.7 Application to French (Lexicon) 18.7.1 Type ------------ ----------- ----------- ---- Attribute Value Example Code ------------ ----------- ----------- ---- Type cardinal deux c

ordinal deuxieme o ------------ ----------- ----------- ----

18.7.2 Gender ------------ ----------- ----------- ---- Attribute Value Example Code ------------ ----------- ----------- ---- Gender masculine premier m

feminine premiere f ------------ ----------- ----------- ----

18.7.3 Number ------------ ----------- ----------- ---- Attribute Value Example Code ------------ ----------- ----------- ---- Number singular premier s

plural premiers p ------------ ----------- ----------- ----

18.7.4 Case Not applicable to French.

18.7.5 Combinations --------- ----------- Tag Example --------- ----------- Mcms- un Mcfs- une Mc-s- zero Mc-p- deux, trois --------- -----------

Recommendations 280 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R Note: Traditionnal grammars usually distinguish un/article and un/numeral. However, it is very difficult to find linguistic tests that enable the two to be discriminated. It is not certain that this distinction will be kept. Cfr. J'ai vu un chat (article) J'ai vu un chat et deux chiens (numeral)

Recommendations 281 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R 18.8 Application to Portuguese Numerals were included as types of adjectives in the Portuguese demo lexicon.

Recommendations 282 May, 1996 EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R 18.9 Application to Danish In traditional dictionaries the word class Numeral covers cardinals and ordinals without any subclassification. Some traditional grammars focus on the adjective-like properties of numerals and treat them as subclasses of the adjective. The EDEMD classifies quantifier, cardinal and ordinal as separate categories. However, the EAGLES attributes are applicable to Danish as well.

18.9.1 Type

Attribute value Example Tag

Type cardinal fem card

ordinal femte ord

Recommendations 283 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R 18.10 Application to Greek Numerals in Greek include both numerals presented in digits (arabic numbers, Greek or Latin letters) and full words. To code this, a specific feature is used (not included in the EAGLES proposal), namely form (see relevant section).

18.10.1 Type Both cardinal and ordinal numerals are recognised in Greek. In addition, we distinguish between multiplicative, analog, fractal and collective numerals.

Attribute value Gr. example Gr. tag

Type cardinal mydhen, 0 NmCd

ordinal tritos, 3os NmOd mult triplos, 3plos NmMl analog dhiplasios NmAn fractal tria tetarta, 3/4 NmFr collect dhekaria NmCt

18.10.2 Gender, Number and Case Given that numerals in Greek act as nouns and/or adjectives, they are further coded for Gender, Number and Case. When they function as adjectives, they must agree in these three features with the noun they refer to.

Not all values of these three features are applicable to numerals, as shown in the following tables. In fact, the values masc-fem for Gender and invariant for Number are not used at all.

Attribute value Gr. example Gr. tag

Gender masculine prwtos NmOdMa

feminine prwty NmOdFe neuter prwto NmOdNe common dhwdheka NmCdCo

Attribute value Gr. example Gr. tag

Number singular prwtos NmOdMaSg

plural prwtoi NmOdMaPl

Recommendations 284 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R

Attribute value Gr. example Gr. tag

Case nom prwtos NmOdMaSgNm

gen prwtou NmOdMaSgGe acc prwto NmOdMaSgAc

l-spec voc prwte NmOdMaSgVo l-spec indcl dhuo NoCdCoPlIc

18.10.3 Function This attribute is used in Greek in order to distinguish between numerals functioning as nouns and numerals that function as adjectives. Adjectival numerals may appear on their own, i.e. the noun they refer to may be ommitted:

Dhwse mou dhwdheka vivlia.

Dhwse mou dhwdheka.

Attribute value It. example It. tag

Function adj dhwdeka NmCdCoPlIcAj

l-spec nom dhwdekadha NmCtCdFeSgNmNo

18.10.4 Form This feature is used to code the form of the numeral (i.e. type of digits or full word-form). All the values are language-specific.

Attribute value Gr. example Gr. tag

Form word dhwdeka NmCdCoPlIcAjWd

l-spec number 12 NmCdCoPlIcAjNu l-spec gr-num ib NmCdCoPlIcAjGn l-spec lat-num XII NmCdCoPlIcAjLn

Recommendations 285 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R 18.11 Application to Swedish 18.11.1 Type

Attribute value Example Tag Type cardinal fem card

ordinal femte ord

Recommendations 286 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R 18.12 Application to Catalan The code AN identifies cardinal numerals. Ordinal numerals are categorized as normal adjectives.

Recommendations 287 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R 18.13 Application to Irish A language-specific feature is the personal form of numerals, i.e. special noun forms to indicate a group of one, two, three, etc. people.

18.13.1 Type ======================================================================== Attribute Value Example Tag ======================================================================== Type Cardin aon, do', tri', ceathair P

Ordinal ce'ad, dara, tri'u', ceathru' P3sm Personal duine, beirt, triu'r, ceathrar P3sf ========================================================================

Recommendations 288 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R 19 Interjections No subcategories are foreseen.

19.1 Application to Italian In the Italian corpus and lexicon, Interjections are tagged I. There are no subcategories.

19.2 Application to German The IMS-Tagset provides the tag ITJ for interjections. There are no subcategories.

19.3 Application to English There are no subcategories of interjections. To avoid confusion over the use of "I", this wordclass is tagged "Ij".

19.4 Application to Dutch CELEX has no distinctive tags for interjections.

19.5 Application to French (Corpus) Although they are a valid class in French, interjections are not coded as such in the IBMF tagset.

19.6 Application to French (Lexicon) In French lexicon, interjections are tagged I.

19.7 Application to Portuguese This category is not included in the Portuguese demo lexicon.

19.8 Application to Danish There are no subcategories in Danish. The EDEMD has no description of interjections.

19.9 Application to Greek In the Greek Morphological Lexicon as well as the Corpus, interjections are coded with the tag "Ij". No subcategories or other features are used for this category.

19.10 Application to Swedish In Swedish there are no subcategorisations for Interjections.

Recommendations 289 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R 19.11 Application to Catalan In catalan corpus and lexicon, Interjections are tagged I. There are no subcategories.

19.12 Application to Irish The tag ITJ is used for interjections.

Recommendations 290 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R 20 Unique membership class 20.1 Comments This category should contain language-specific phenomena.

EAGLES recommends keeping this class as small as possible. All the members of this class are to be seen at the level 2b of language-specific distinctions, e.g. `infinitive marker' for English, German, Dutch, Danish), `existential' for English, etc.

AlethDic has a special class "Mot not autonome", in which e.g. hui is included.

20.2 Application to German Categories with unique members are particles like negation (nicht ), infinitive marker (zu), superlative marker (am). The IMS-tagset includes another particle type for separable prefixes. It covers prefixes like an, aus, ein, ... (mostly identical to prepositions), but also other (non-adverbial) prefixes like rad, statt, instand .

Example: er kommt an/PTKVZ

er kauft ein/PTKVZ er f"ahrt rad/PTKVZ (not: Rad !) es findet statt/PTKVZ er h"alt das Haus instand/PTKVZ

20.3 Application to English 20.3.1 Particle-Type

Attribute values Examples Tags Particle-Type infinitive to (+ Infin.) UI

negative not, n't UN existential there (is/are) UX

20.4 Application to Dutch We have recently sent to the EAGLES group of Professor G. Leech the proposal of attributing a value for Dutch in this group: infinitive-marker to mark up the prepositions `om' and `te' and `om te'.

20.5 Application to French (Corpus and Lexicon) We see no language specifics that would be coded in the IBMF tagset and would not fit in other EAGLES classes.

Recommendations 291 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R 20.6 Application to Greek This class could be used for the codification of particles, which form a distinct category in Greek. The only feature pertinent to this category is that of Type.

20.6.1 Particle-Type Particles can be used to form compound forms of tense (e.g. future) or mood (e.g. subjunctive), or as introducers of negation, or for suggesting, swearing, or wishing.

Attribute value Gr. Example Gr. Tag Type fut tha PtFu

neg dhen PtNg subj na PtSj other as, makari, eithe PtOt

The value other is used to group together various functions of particles, such as hesitating, wishing, etc. which are semantically motivated and cannot be distinguished unless the linguistic and extra-linguistic context is taken into account.

20.7 Application to Irish ======================================================================== Attribute Value Example Tag ======================================================================== Particle-type infinitive a, a' PARINF

negative ni' PARNEG vocative a PARVOC subj go PARSUBJ ========================================================================

Recommendations 292 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R 21 Residual

RESIDUAL Type Gend Numb Function

MULTILEX Misc*

GENELEX affixes

shortcuts fused forms comp. forms AlethDic

NERC letters

symbols formulae abbreviations Foreign Words* Leech foreign words m s

alphabetic symbols f p acronyms n formulae abbreviations unclassified

EAG-L0 RESIDUAL EAG-L1 foreign words m sg

alphabetic symbols f pl formulae n acronyms abbreviations unclassified affixes shortcuts fused forms trunc.forms comp. forms EAG-L2a Noun-c

Noun-p Adj EAG-L2b c n

Recommendations 293 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R 21.1 Comments EAGLES recommends keeping this class as small as possible and the specification of links to other classes when they are clear. Some annotation practices prefer to treat phenomena included here as belonging to other parts of speech (e.g. foreign words treated as nouns having number and gender).

21.1.1 Type MULTILEX proposes two indistinct classes Misc and None for words for which it is unclear what category should be assigned and for words to which no category should be assigned. It can be argued that these two categories should be mapped onto the Residual class proposed here.

In NERC Foreign Words were kept apart from the so-called Rest Group and treated as a separate category, since the former may behave in a different way syntactically.

Recommendations 294 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R 21.2 Application to Italian 21.2.1 Type

Attribute values Examples Tags Type foreign word polis FW

abbreviation Sig. abbr sigla CNR sigla

21.3 Application to German Only punctuation is included in the IMS-tagset; no tags are provided yet for symbols, formula, foreign words etc.

Abbreviations are classified according to the "full" word form, and are marked by an additional feature ABK. Acronyms are generally classified as proper names.

21.3.1 Type

Attribute values Examples Tags Type punctuation ? IPNORM

abbreviation Tel. NN:ABK

dt. ADJA:ABK acronyms USA NE:ABK

21.4 Application to English 21.4.1 Type

Attribute values Examples Tags Type foreign word mawashi RFW

symbol $, j RSY formula X/21= RFO unclassified la-la-la RUN

21.4.2 Number

Attribute values Examples Tags Number singular A, b RSYs

plural As, b's RSYp

21.5 Application to Dutch We have recently sent to the EAGLES group of Professor G. Leech the proposal of attributing an additional value for Dutch sub Residual-Type: Acronym.

Recommendations 295 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R 21.6 Application to French (Corpus) The IBMF tagset codes punctuation marks as specific tags. Indeed, they are morphological manifestations and can help predict other tags. The tagset distinguishes weak punctuation (tag AAAA), strong punctuation (YAAA) and sentence boundary (ZTRM).

21.7 Application to French (Lexicon) --------- ----------- Tag Example --------- ----------- X IBM --------- -----------

21.8 Application to Greek 21.8.1 Foreign Word In the Greek Morphological Lexicon, we currently keep apart foreign words from the rest of the elements included in EAGLES under the general category of "residual". This distinction allows us to code further information on foreign words necessitated by the morpho-syntactic system of Greek. In this category, we include only those foreign words that have not been adapted to the morphological system of Greek and, furthermore, are not considered "Greek" words.

However, the inclusion of foreign words under the "Residual" category can be easily performed, given the tag "Fw" used for this category.

Two additional features are used for foreign words: form and fw-cat. The first attribute is used to distinguish between foreign words that are transliterated in the Greek alphabet and those kept in their original Latin form, and the second attribute is used to code the grammatical category of the foreign word in order to facilitate further syntactic parsing of the texts.

Attribute value Gr. example Gr. tag

Form translit Tzeikomp FwTr

original Jacob FwOr

Attribute value Gr. example Gr. tag

Fw-cat noun Tzeikomp FwTrNo

abbrev EEC FwOrAb other cogito FwOrOt

Recommendations 296 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R 21.8.2 Type In this section, we present all other elements included in the EAGLES Residual category. These are coded in the ILSP Morphological Lexicon as belonging to the category "Rest-group". The basic feature for these elements is that of Type.

Attribute values Examples Tags Type foreign word Tzeikomp Fw

abbr forol. RgAb formula F(x) RgFo symbol OE RgSy l-spec date (17/03/92) RgDa

No distinction is made at present between abbreviations and acronyms, while the extra value of date is used for the codification of dates, either written in digits or in letters.

21.9 Application to Catalan No-part-of-speech codes like SIG (for acronyms), NC (non-codified) are present in the corpus that could be treated as "residual" information.

21.10 Application to Irish 21.10.1 Type ======================================================================== Attribute Value Example Tag ======================================================================== Type foreign word Glasnost FW

abbreviation srl., lch., [etc., p. (page)] ABB symbol 20.75p, x (y + 3) SYM toponym Baile Atha Cliath [Dublin] TOPONM acronym SAM [USA] ACRONM ========================================================================

Recommendations 297 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R 22 References Aglamissis, Y., M. Gavrilidou, P. Labropoulou, H. Papageorgiou, S. Piperidis and C. Raptis (1994): "The ILSP Morphological Tagset". NERC-2 Document.

Bel N., Calzolari N., Monachini M. (Coords.) (1995): "Common Specifications and Notation For Lexicon Encoding", MULTEXT D-1.6-B Deliverable.

Bel N., M. Villegas (1993): "Morphosyntax for Spanish", EAGLES Input Document, Barcelona. Bindi R., M. Monachini, P. Orsolini (1991): "Italian Reference Corpus" NERC Technical Report, ILC, Pisa.

Burnage G. (1990): CELEX. A guide for users. Calzolari, N. (1994), "European Efforts Towards Standardizing Language Resources", in P. Steffens (ed.), Machine Translation and Lexicon (Springer).

Calzolari N., Ceccotti M.L., Roventini A. (1983): "Documentazione sui tre nastri contenenti il DMI", ILC Technical Report, Pisa.

Calzolari N., Monachini M. (1995): "EAGLES Proposal for Morphosyntactic Standards: in View of a Ready-to-use Package", to appear in Literary and Linguistic Computing, OUP.

Calzolari, N. and Zampolli, A. (1994), "Standards to Make Natural Language Resources Shareable Resources", in Proceedings of the International Workshop on Shareable Natural Language Resources, Nara.

Danielsson P. and J"arborg J. (1995) "Morphosyntactic description of Swedish language", EAGLES input report.

Dutilh-Ruitenberg T. (1994): "A Comparative Report on Morphosyntactic Categories in Dutch as encoded in the CELEX Dutch Lexical Databases. Augmented with ten proposals for Dutch, EAGLES Input Document, Leiden.

GENELEX Consortium (Apr 1993): "Couche Morphologique", Version 3.0, ASSTRIL, Gsi-Erli, IBM France, SEMA GROUP.

GENELEX Consortium (Sept 1993): "Couche Syntaxique. Les Unites Syntaxique Simple", Tome 1, Version 3.0, ASSTRIL, Gsi-Erli, IBM France, SEMA GROUP.

Gsi-Erli (Jun 1993): "Le Dictionnaire AlethDic", Gsi-Erli, Paris. Guerreiro, P. (1994): "Morphosyntactic Phenomena Encoded in a Portuguese NLP Lexicon. Application of the EAGLES Specificications", EAGLES input document, Lisbon.

Recommendations 298 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R Guerreiro, P. (ed.), (1994): "Linguistic Specifications for Portuguese Computational Lexica", Final Report of GENELEX-PT, Eureka: EU 524.

Heid, U., McNaught, J. (eds.) (1991): "Eurotra-7 Study: Feasibility and Project Definition Study on the Reusability of Lexical and Terminological Resources in Computerised Applications". Eurotra-7 Final Report, Stuttgart.

Heylen, D. (April 1994): "Comments on a Common tagset", EAGLES Input Document, Utrecht.

Lang'e, J.M. (March 1994): "Application of EAGLES morphosyntactic guidelines to a French tagset", EAGLES Input Document, IBM, Paris.

Leech, G. (1992): "Corpus Annotation Schemes", paper presented at the Pisa Corpus Workshop (24-26 January 1992), to be published in Literary and Linguistic Computing, OUP.

Leech G., A. Wilson (1993a): "Invitation Draft", Draft EAGLES Document, Lancaster. Leech G., A. Wilson (1993b): "Tagset Guidelines", Draft EAGLES Document, Lancaster. Leech G., A. Wilson (1994a): "MSAL21", Draft EAGLES Document, Lancaster. Leech G., A. Wilson (1994b): "A Morphosyntactic Tagset for English, making use of the guidelines of the Pisa Document", EAGLES Input Document, Lancaster.

Mac Aogain E. (1995): "Application to Irish of the EAGLES proprosal", EAGLES input document.

Mantzari, E. and M. Gavrilidou (1995): "Report on the Classification of functional words in Greek".

Monachini (1992): "Core set of PoS tags for Italian", ILC Technical Report, Pisa. Monachini, M. (1995), "ELM-IT A typed Italian Incarnation of the EAGLES-TS. Definition of Lexicon Specifications and Classification Guidelines", Pisa.

Monachini M. (ed.) (1995b): "Common Specifications and Notation for Lexicon Encoding of Eastern Languages", Deliverable D1.1 COP Project 106 MULTEXT-East.

Monachini M. (ed.) (1996): "Lexicon: Morphosyntactic Specifications and Language Specific Instantiations", MLAP-PAROLE Deliverable WP4.2.2, Pisa.

Monachini M., N. Calzolari (1994): "Application of EAGLES Proposal for Morphosyntactic encoding to Italian Lexicon and Corpus", EAGLES Input Document, ILC, Pisa.

Monachini M., A. "Ostling (1992a): "Morphosyntactic Corpus Annotation - A Comparison of Different Schemes", NERC-WP8-60, Pisa.

Recommendations 299 May, 1996

EAGLES Morphosyntactic Phenomena EAG-CLWG-MORPHSYN/R Monachini M., A. "Ostling (1992b): "Towards a Minimal Standard for Morphosyntactic Corpus Annotation", NERC-WP8-61, Pisa.

MULTILEX Consortium (Apr 1993): "Standards for Multifunctional Lexicon", CAP GEMINI, Philips, Univ. of Surrey, Univ. of Bochum, Univ. of Muenster.

NERC 1994, "NERC Final Report", (to appear), Giardini Editori, Pisa. Schiller A. (1993): "Tagsets f"ur Deutsch", Draft EAGLES input Document, Stuttgart. Schiller A. (1994): "Guidelines f"ur das Tagging deutscher Textcorpora (Kleines und erweitertes Tagset)", Univ. Stuttgart, Internal Document.

Schiller A., Thielen C. (forthc.): "Ein kleines und erweitertes Tagset f"urs Deutsche", to appear in: Tagungsberichte des Arbeitstreffens Lexikon + Text, 17./18. Februar 1994, Schloss Hohent"ubingen, Niemeyer: Lexicographica Series Maior, T"ubingen, Spring 1995.

Soler J. (1995): "Application of the EAGLES Proposal to Catalan", EAGLES input report. TEI AI 1W2 (June 1991): "List of Common Morphological Feature for Inclusion in TEI Starter Set of Grammatical-Annotation Tags".

Teufel, S. (1995), "A Typed German Incarnation of the EAGLES-TS. Definition of Lexicon Specifications and Classification Guidelines", Stuttgart.

Teufel, S. (1995), "ELM-EN: EAGLES Validation (WP4) Task on Tagset Mapping"), Stuttgart. Veronis J., Khouri L., Meunier C. (1994a): "Application of the EAGLES-L1 Proposal to French", Draft EAGLES input Document, Aix-en-Provence.

Veronis J., Khouri L., Meunier C. (1994): "Proposal for Morphosyntactic Encoding in MULTEXT", Aix-en-Provence.

Von Rekowsky, U. (1995), "A French Incarnation of the EAGLES-TS. Definition of Lexicon Specifications and Classification Guidelines", Paris.

Recommendations 300 May, 1996