An End-to-End Discriminative Approach to Machine Translation

Subscribe to email list

Please select the email list(s) to which you wish to subscribe.

You are here

An End-to-End Discriminative Approach to Machine Translation

TitleAn End-to-End Discriminative Approach to Machine Translation
Publication TypeConference Paper
Year of Publication2006
AuthorsLiang, P, Bouchard-Cote, A, Klein, D, Taskar, B
Conference NameCOLING/ACL 2006, VOLS 1 AND 2, PROCEEDINGS OF THE CONFERENCE
PublisherAssoc Computat Linguist
Conference Location209 N EIGHTH STREET, STROUDSBURG, PA 18360 USA
ISBN Number978-1-932432-65-7
AbstractWe present a perceptron-style discriminative approach to machine translation in which large feature sets can be exploited. Unlike discriminative reranking approaches, our system can take advantage of learned features in all stages of decoding. We first discuss several challenges to error-driven discriminative approaches. In particular, we explore different ways of updating parameters given a training example. We find that making frequent but smaller updates is preferable to making fewer but larger updates. Then, we discuss an array of features and show both how they quantitatively increase BLEU score and how they qualitatively interact on specific examples. One particular feature we investigate is a novel way to introduce learning into the initial phrase extraction process, which has previously been entirely heuristic.