Paper: A Discriminative Framework For Bilingual Word Alignment

ACL ID H05-1011
Title A Discriminative Framework For Bilingual Word Alignment
Venue Conference on Empirical Methods in Natural Language Processing
Session Main Conference
Year 2005
Authors

Bilingual word alignment forms the foun- dation of most approaches to statistical machine translation. Current word align- ment methods are predominantly based on generative models. In this paper, we demonstrate a discriminative approach to training simple word alignment mod- els that are comparable in accuracy to the more complex generative models nor- mally used. These models have the the advantages that they are easy to add fea- tures to and they allow fast optimization of model parameters using small amounts of annotated data. 1 Motivation Bilingual word alignment is the first step of most current approaches to statistical machine translation. Although the best performing systems are “phrase- based” (e.g, Och and Ney, 2004), possible phrase translations are normally first extracted...