Etude de l'impact d'un lexique bilingue spécialisé sur la performance d'un moteur de traduction à base d'exemples (Studying the impact of a specialized bilingual lexicon on the performance of an example-based machine translation engine)[In French].

Studying the impact of a specialized bilingual lexicon on the performance of an example-based machine translation engine Non-availability of parallel corpora for several languages and for specific domains is a major challenge for domain adaptation in statistical machine translation. We present, in this paper, an Example-Based Machine Translation engine based on cross-language information retrieval and which needs only a monolingual corpus in the target language. In particular, we investigate the impact of using a domain-specific bilingual lexicon on the performance of this translation engine. We evaluate and compare this translation engine to the statistical machine translation system Moses using English-French parallel corpora Europarl (European Parliament Proceedings) and Emea (European Medicines Agency Documents). The obtained results show that the BLEU score of the Example-Based Machine Translation engine is close to the ones of Moses for the Europarl corpus and better for the Emea corpus. MOTS-CLÉS : Traduction automatique, recherche d’information interlingue, lexique bilingue, modèle de traduction, modèle de langue, automate d’états finis, champs conditionnels aléatoires.