Example-Based Paraphrasing for Improved Phrase-Based Statistical Machine Translation

Proceedings of the 2010 Conference on Empirical Methods in Natural Language Processing, pages 656–666, MIT, Massachusetts, USA, 9-11 October 2010. c⃝2010 Association for Computational Linguistics Example-based Paraphrasing for Improved Phrase-Based Statistical Machine Translation Aur´elien Max LIMSI-CNRS & Univ. Paris Sud Orsay, France aurelien.max@limsi.fr Abstract In this article, an original view on how to improve phrase translation estimates is pro- posed. This proposal is grounded on two main ideas: first, that appropriate examples of a given phrase should participate more in build- ing its translation distribution; second, that paraphrases can be used to better estimate this distribution. Initial experiments provide ev- idence of the potential of our approach and its implementation for effectively improving translation performance. 1 Introduction Phrase translation estimation in Statistical Phrase- based Translation (Koehn et al., 2003) is hampered by the availability of both too many and too few training instances. Recent results on tera-scale SMT (Lopez, 2008) show that access to many training examples1 can lead to significant improvements in translation quality. Also, providing indirect train- ing instances via synonyms or paraphrases for pre- viously unseen phrases can result in gains in trans- lation quality, which are more apparent when little training data is originally available (Callison-Burch et al., 2006; Marton et al., 2009; Mirkin et al., 2009; Aziz et al., 2010). Although there is a consensus on the importance of using more parallel data in SMT, it has never been formally shown that all additional training instances are actually useful in predicting contextually appropriate translation hypotheses. 1To be more accurate, works such as that of (Lopez, 2008) have recourse to random sampling to build models of a manage- able size in a reasonable amount of time. Attempts at limiting training parallel sentences to those resembling test data through thematic adapta- tion (Hildebrand et al., 2005) indeed confirm that large quantities of training data cannot compen- sate for the requirement for contextually appropriate training instances. In fact, it is important that phrase translation models adequatly reflect contextual pref- erences for each phrase occurrence in a text. A vari- ety of recent works have used dynamically adapted translation models, where each phrase occurrence has its own translation distribution (Carpuat and Wu, 2007; Stroppa et al., 2007; Max et al., 2008; Gim- pel and Smith, 2008; Haque et al., 2009) derived from local contextual information in the training ex- amples.2 These approaches are supported by the study of (Wisniewski et al., 2010) which shows that phrase-based SMT systems are expressive enough to achieve very high translation performance and there- fore suggests a better scoring of phrases. The apparent tradeoff between the number of training examples and their appropriateness in each indivual context naturally asks for means of increas- ing the number of appropriate examples. Exploiting comparable corpora for acquiring translation equiva- lents (Munteanu and Marcu, 2005; Abdul-Rauf and Schwenk, 2009) offers interesting prospects to this issue, but so far focus has not been so much on con- text appropriateness as on globally increasing the number of biphrase examples. 2The …

Content truncated to 500 words