Statistical Post-Editing of Machine Translation for Domain Adaptation

Raphaël Rubino (1) , Stéphane Huet (1) , Fabrice Lefèvre (1) , Georges Linarès (1)

1 LIA - Laboratoire Informatique d'Avignon

This paper presents a statistical approach to adapt out-of-domain machine translation systems to the medical domain through an unsupervised post-editing step. A statistical post-editing model is built on statistical machine translation (SMT) outputs aligned with their translation references. Evaluations carried out to translate medical texts from French to English show that an out-of-domain machine translation system can be adapted a posteri-ori to a specific domain. Two SMT systems are studied: a state-of-the-art phrase-based implementation and an online publicly available system. Our experiments also indicate that selecting sentences for post-editing leads to significant improvements of translation quality and that more gains are still possible with respect to an oracle measure.

Domaines: Informatique et langage [cs.CL]

Raphaël Rubino, Stéphane Huet, Fabrice Lefèvre, Georges Linarès. Statistical Post-Editing of Machine Translation for Domain Adaptation. 16th Annual Conference of the European Association for Machine Translation (EAMT), May 2012, Trento, Italy. ⟨hal-01320242⟩