Word-level confidence estimation for machine translation using phrase-based translation models

Abstract

Confidence measures for machine translation is a method for labeling each word in an automatically generated translation as correct or incorrect. In this paper, we will present a new approach to confidence estimation which has the advantage that it does not rely on system output such as Nbest lists or word graphs as many other confidence measures do. It is, thus, applicable to any kind of machine translation system. Experimental evaluation has been performed on translation of technical manuals in three different language pairs. Results will be presented for different machine translation systems to show that the new approach is independent of the underlying machine translation system which generated the translations. To the best of our knowledge, the performance of the new confidence measure is better than that of any existing confidence measure. © 2005 Association for Computational Linguistics.