Relations sémantiques en traduction automatique : degrés d’inclusion et limites d’application

Statistical machine translation (SMT) often fails when faced with the problem of recognition of semantic relations in context. This article aims to give an overview and to detect possible reasons of semantic incoherence in translations. We try to demonstrate what effect the consideration of linguistic and natural language processing (NLP) methods can have on the quality of translations. We try to argue in favour of the introduction of word sense disambiguation (WSD) methods in the word selection process.