Alignement de traductions rares à l'aide de paires de phrases non alignées

Abstract : There generally exist numerous ways to translate a word or a phrase in another language. Among these translations, some are very common while others are far less so, according to a zipfian law. As with the rest of the world, translation memories are googlized, leading to poorly handled or even simply ignored rare translations, while they are often of good quality. In this paper, we tackle this problem in a transpotting framework. We show that these rare translations are harder to identify than common translations. We describe an original approach based on the word alignment of sentences which are not aligned. We show that this approach significantly improves the identification of those rare translations.
Document type :
Conference papers
Complete list of metadatas

Cited literature [13 references]  Display  Hide  Download
Contributor : Stéphane Huet <>
Submitted on : Friday, February 15, 2019 - 6:54:00 PM
Last modification on : Wednesday, April 17, 2019 - 12:15:36 PM
Long-term archiving on : Friday, May 17, 2019 - 10:00:25 AM


Files produced by the author(s)


  • HAL Id : hal-02021394, version 1


Julien Bourdaillet, Stéphane Huet, Philippe Langlais. Alignement de traductions rares à l'aide de paires de phrases non alignées. 17ème conférence sur le Traitement Automatique des Langues Naturelles (TALN), 2010, Montréal, Canada. ⟨hal-02021394⟩



Record views


Files downloads