Presentation is loading. Please wait.

Presentation is loading. Please wait.

On the status of the deep-syntactic structure Sylvain Kahane Lattice, Paris 7 / Paris 10 MTT 2003, ENS 18 juin 2003.

Similar presentations


Presentation on theme: "On the status of the deep-syntactic structure Sylvain Kahane Lattice, Paris 7 / Paris 10 MTT 2003, ENS 18 juin 2003."— Presentation transcript:

1

2 On the status of the deep-syntactic structure Sylvain Kahane Lattice, Paris 7 / Paris 10 MTT 2003, ENS 18 juin 2003

3 Introduction n The deep-syntactic structure [DSyntS]: –plays a central in the theory (paraphrasing and lexical functions) –is one of the less defined representations “A deep-syntactic relation represents a FAMILY OF SYNTACTIC CONSTRUCTIONS of the same structural type, regardless of their semantic content” (Mel'cuk 1988) –Peter spoke [to Mary] II –Peter gave a book [to Mary] III

4 Introduction n I think: 1.It is useless to consider the DSyntS as an intermediate step in the semantics-syntax correspondence 2.It is useful to consider such a structure for paraphrasing and LF encoding. n How to solve this apparent paradox?

5 To solve the paradox n The solution comes from TAG (Joshi 1987, Vijay-Shanker 1987) : –the structure of a sentence can be computed from the combination of elementary structures associated to the words (or idioms) –The process of combination can be stored in a derivation tree, comparable to the MTT deep-syntactic tree (Rambow & Joshi 1992)

6 To solve the paradox (2) n Therefore, we can obtain a structure equivalent to DSyntS without presupposing it! –Fundamental theoretical role: how the linguistic signs combine to yield a sentence –Central role in paraphrasing: replacing a combination of signs by another one

7 Plan n Introduction n Semantics-syntax correspondence and derivation structure n Idioms and lexical functions (for inflectional morphems see Kahane 2002 and TALN 2003) n Semantics-syntax mismatches n Conclusion

8 Semantics-syntax correspondence n Direct correspondence between a semantic graph and a surface-syntactic dependency tree n Rules = elementary structures associated to morphems (lexemes, LFs, inflectional morphems, “constructions”) n Rules are combined by unification (  superimposition)

9 ‘speak’ 12 Lexical morphems subjiobj PARLER (V)  v,  m sem: ‘speak’ arg1: x arg2: y À (Prep) ¬sém prep (N) sém: y (N) sém: x subj iobj À (Prep) prep (N) PARLER (V)  v,  m

10 Quasi-dependency Pierre essaye de dormir ‘Peter tries to sleep’

11 Quasi-dependency Pierre essaye de dormir ‘Peter tries to sleep’

12 A remark n An important difference between our rules and traditional MTT correspondence rules: our rules are rules of potential correspondence

13 Inflectional morphems n mood = finiteness (finite, inf, past-part, pres-part) + true-mood (ind, subj, imp) n indicative introduces a tense request

14 Correspondence and derivation Zoé essaye de manger la pomme ‘Zoe tries to eat the apple’

15 12 ESSAYER ZOÉ MANGER POMME ind,present,actif def,sg 2 actif 1 derivation structure = deep syntax syntax derived structure semantics

16 Conventions for the derivation structure

17 Idioms PRENDRE LE TAUREAU PAR LES CORNES ‘take the bull by the horns’ = ‘face up’

18 Lexical functions À LA FOLIE ‘to the madness’ Magn is a generalization of such signs

19 Lexical functions

20 Encoding of LFs and derivation structure

21 Semantics-syntax mismatches n Tough-movement: a book easy to read n a book easy to ask to read *un livre facile à demander de lire n un livre facile à lire il est facile de lire ce livre = ‘a book such that read this book is easy’

22

23

24 Tough-movement un livre facile à lire = ‘un livre tel que lire ce livre et facile’

25 Tough-movement un livre facile à lire = ‘un livre tel que lire ce livre et facile’

26 lire ce livre (est) facile (il est) facile de lire ce livre impersonnel ce livre (est) facile à lire Adj à Vinf

27 Conclusion n Avantages du formalisme –une seule structure pour un grand nombre d’emplois d’un lexème (cf. CG ou TAG) –pas de description syntagmatique (cf. HPSG)  pas de nécessité de saturer un syntagme pour combiner avec sa tête n Structure de dérivation non arborescente (à cause des quasi-dépendances)

28

29 Thank you for your attention


Download ppt "On the status of the deep-syntactic structure Sylvain Kahane Lattice, Paris 7 / Paris 10 MTT 2003, ENS 18 juin 2003."

Similar presentations


Ads by Google