Using Paraphrases for Parameter Tuning in Statistical Machine Translation

DMCA

Using Paraphrases for Parameter Tuning in Statistical Machine Translation

Cached

Download Links

by Nitin Madnani , Necip Fazil Ayan , Philip Resnik , Bonnie J. Dorr

Citations:

41 - 10 self

BibTeX

@MISC{Madnani_usingparaphrases,
    author = {Nitin Madnani and Necip Fazil Ayan and Philip Resnik and Bonnie J. Dorr},
    title = {Using Paraphrases for Parameter Tuning in Statistical Machine Translation},
    year = {}
}

OpenURL

Abstract

Most state-of-the-art statistical machine translation systems use log-linear models, which are defined in terms of hypothesis features and weights for those features. It is standard to tune the feature weights in order to maximize a translation quality metric, using held-out test sentences and their corresponding reference translations. However, obtaining reference translations is expensive. In this paper, we introduce a new full-sentence paraphrase technique, based on English-to-English decoding with an MT system, and we demonstrate that the resulting paraphrases can be used to drastically reduce the number of human reference translations needed for parameter tuning, without a significant decrease in translation quality. 1

Keyphrases

parameter tuning statistical machine translation translation quality new full-sentence paraphrase technique held-out test sentence corresponding reference translation mt system hypothesis feature log-linear model feature weight significant decrease human reference translation reference translation state-of-the-art statistical machine translation system english-to-english decoding

DMCA