Skip to Main content Skip to Navigation
Theses

Analyse d'énoncés oraux pour le dialogue homme-machine à l'aide de grammaires lexicalisées d'arbres

Abstract : This thesis investigates a linguistically motivated grammatical formalism for the robust parsing of spoken utterances in the context of Man-Machine Dialogue. We propose a framework for the spoken language processing based on Lexicalized Tree Adjoining Grammars (LTAG). LTAG is an interesting formalism for both parsing and linguistic descriptions. In order to adapt the formalism to spoken dialogue, we present first a parsing algorithm that delivers derivations for all the most extended grammatical chunks of an utterance - the connection driven parsing algorithm. The efficiency of this algorithm is based on tabular techniques and structure sharing precompilations. Complementary, we propose principles to capture spoken language constraints in a classical LTAG and additional parsing repairs rules dedicated to phenomena as spoken disfluencies. These propositions are implemented in a system called EGAL (Lexicalized Tree Grammars Extractions) which allows an assisted specialization of a general grammar to an application sublanguage according to a Wizard of Oz corpora. The system allows also to test various parsing algorithms in order to evaluate the resulting grammar and our robust algorithms.
Document type :
Theses
Complete list of metadatas

https://hal.univ-lorraine.fr/tel-01747108
Contributor : Thèses Ul <>
Submitted on : Thursday, March 29, 2018 - 10:51:11 AM
Last modification on : Tuesday, April 24, 2018 - 1:52:08 PM
Long-term archiving on: : Thursday, September 13, 2018 - 1:17:57 PM

File

SCD_T_1999_0259_LOPEZ.pdf
Files produced by the author(s)

Identifiers

  • HAL Id : tel-01747108, version 1

Collections

Citation

Patrice Lopez. Analyse d'énoncés oraux pour le dialogue homme-machine à l'aide de grammaires lexicalisées d'arbres. Autre [cs.OH]. Université Henri Poincaré - Nancy 1, 1999. Français. ⟨NNT : 1999NAN10259⟩. ⟨tel-01747108⟩

Share

Metrics

Record views

107

Files downloads

45