Beyond log-linear models

Kevin Duh, Katrin Kirchhoff
2008 Proceedings of the 46th Annual Meeting of the Association for Computational Linguistics on Human Language Technologies Short Papers - HLT '08   unpublished
Current re-ranking algorithms for machine translation rely on log-linear models, which have the potential problem of underfitting the training data. We present BoostedMERT, a novel boosting algorithm that uses Minimum Error Rate Training (MERT) as a weak learner and builds a re-ranker far more expressive than log-linear models. BoostedMERT is easy to implement, inherits the efficient optimization properties of MERT, and can quickly boost the BLEU score on N-best re-ranking tasks. In this paper,
more » ... we describe the general algorithm and present preliminary results on the IWSLT 2007 Arabic-English task.
doi:10.3115/1557690.1557701 fatcat:6d7rnqemlngarj5l3wfsprcf7i