Linguistic-Based Computational Treatment of Textual Entailment Recognition

Abstract : In this thesis, I investigate how lexical resources based on the organisation of lexical knowledge in classes which share common (syntactic, semantic, etc. features support natural language processing and in particular symbolic recognition of textual entailment. First, I present a robust and wide coverage approach to lexico-structural verb paraphrase recognition based on Levin's (1993) classification of English verbs. Then, I show that by extending Levin's framework to general inference patterns, a classification of English adjectives can be obtained that compared with previous approaches, provides a more fine grained semantic characterisation of their inferential properties. Further, I develop a compositional semantic framework to assign a semantic representation to adjectives based on an ontologically promiscuous approach (Hobbs, 1985) and thereby supporting first order inference for all types of adjectives including extensional ones. Finally, I present a test suite for adjectival inference I developed as a resource for the evaluation of computational systems handling natural language inference.
Document type :
Theses
Complete list of metadatas

Cited literature [105 references]  Display  Hide  Download

https://hal.univ-lorraine.fr/tel-01748535
Contributor : Thèses Ul <>
Submitted on : Thursday, March 29, 2018 - 11:37:38 AM
Last modification on : Tuesday, April 24, 2018 - 1:53:21 PM
Long-term archiving on : Friday, September 14, 2018 - 7:22:48 AM

File

SCD_T_2008_0099_AMOIA.pdf
Files produced by the author(s)

Identifiers

  • HAL Id : tel-01748535, version 1

Collections

Citation

Marilisa Amoia. Linguistic-Based Computational Treatment of Textual Entailment Recognition. Other [cs.OH]. Université Henri Poincaré - Nancy 1, 2008. English. ⟨NNT : 2008NAN10099⟩. ⟨tel-01748535⟩

Share

Metrics

Record views

37

Files downloads

40