Skip to Main content Skip to Navigation
Theses

La détection et la traduction automatiques de l'ellipse : enjeux théoriques et pratiques

Abstract : This thesis examines the automatic processing of elliptical phenomena in English. Situated at the crossroads of various disciplines, such as theoretical linguistics, corpus linguistics, computational linguistics and translation studies, it implements an experimental approach in order to achieve two main goals. The first one consists in establishing that ellipsis can be detected automatically, and the second one in exploring procedures to facilitate its machine translation from English into French.For these purposes, this thesis resorts to morpho-syntactic analyses, which seem sufficient for the automatic detection of some types of ellipsis, since through breaking down the phenomenon into its specific features it makes it possible to identify it among others. The investigation draws on a parallel, multi-genre corpus compiled to test the research hypotheses. In order to develop detection patterns and exploit the corpus, the CoreNLP package developed at the University of Stanford (US) is used, and its limitations in the processing of ellipsis are highlighted. Findings and results emphasize the close link between the detection of ellipsis and its machine translation as a key factor in understanding translation errors caused by its automatic processing.
Document type :
Theses
Complete list of metadata

Cited literature [207 references]  Display  Hide  Download

https://tel.archives-ouvertes.fr/tel-02526670
Contributor : ABES STAR :  Contact
Submitted on : Tuesday, March 31, 2020 - 4:59:10 PM
Last modification on : Friday, October 23, 2020 - 4:45:41 PM

File

Hamza_Anissa_2019_ED520.pdf
Version validated by the jury (STAR)

Identifiers

  • HAL Id : tel-02526670, version 1

Collections

Citation

Anissa Hamza. La détection et la traduction automatiques de l'ellipse : enjeux théoriques et pratiques. Linguistique. Université de Strasbourg, 2019. Français. ⟨NNT : 2019STRAC015⟩. ⟨tel-02526670⟩

Share

Metrics

Record views

114

Files downloads

298