Minimum error rate training in statistical machine translation

TLDR

Statistical machine translation models are typically trained with maximum likelihood, yet this criterion has a weak link to actual translation quality on unseen data. The study analyzes training criteria that directly optimize translation quality and introduces an efficient algorithm for unsmoothed error count training. The authors employ recent automatic evaluation metrics to guide training and implement the unsmoothed error count algorithm. They demonstrate that directly incorporating the final evaluation criterion during training yields significantly better translation results.

Abstract

Often, the training procedure for statistical machine translation models is based on maximum likelihood or related criteria. A general problem of this approach is that there is only a loose relation to the final translation quality on unseen text. In this paper, we analyze various training criteria which directly optimize translation quality. These training criteria make use of recently proposed automatic evaluation metrics. We describe a new algorithm for efficient training an unsmoothed error count. We show that significantly better results can often be obtained if the final evaluation criterion is taken directly into account as part of the training procedure.