Learning a Phrase-based Translation Model from Monolingual Data with Application to Domain Adaptation

Proceedings of the 51st Annual Meeting of the Association for Computational Linguistics, pages 1425–1434, Sofia, Bulgaria, August 4-9 2013. c⃝2013 Association for Computational Linguistics Learning a Phrase-based Translation Model from Mon- olingual Data with Application to Domain Adaptation Jiajun Zhang and Chengqing Zong National Laboratory of Pattern Recognition Institute of Automation, Chinese Academy of Sciences, Beijing, China {jjzhang, cqzong}@nlpr.ia.ac.cn Abstract Currently, almost all of the statistical ma- chine translation (SMT) models are trained with the parallel corpora in some specific domains. However, when it comes to a lan- guage pair or a different domain without any bilingual resources, the traditional SMT loses its power. Recently, some research works study the unsupervised SMT for in- ducing a simple word-based translation model from the monolingual corpora. It successfully bypasses the constraint of bitext for SMT and obtains a relatively promising result. In this paper, we take a step forward and propose a simple but effec- tive method to induce a phrase-based model from the monolingual corpora given an au- tomatically-induced translation lexicon or a manually-edited translation dictionary. We apply our method for the domain adaptation task and the extensive experiments show that our proposed method can substantially improve the translation quality. 1 Introduction During the last decade, statistical machine trans- lation has made great progress. Novel translation models, such as phrase-based models (Koehn et a., 2007), hierarchical phrase-based models (Chiang, 2007) and linguistically syntax-based models (Liu et a., 2006; Huang et al., 2006; Gal- ley, 2006; Zhang et al, 2008; Chiang, 2010; Zhang et al., 2011; Zhai et al., 2011, 2012) have been proposed and achieved higher and higher translation performance. However, all of these state-of-the-art translation models rely on the parallel corpora to induce translation rules and estimate the corresponding parameters. It is unfortunate that the parallel corpora are very expensive to collect and are usually not available for resource-poor languages and for many specific domains even in a resource-rich language pair. Recently, more and more researchers concen- trated on taking full advantage of the monolin- gual corpora in both source and target languages, and proposed methods for bilingual lexicon in- duction from non-parallel data (Rapp, 1995, 1999; Koehn and Knight, 2002; Haghighi et al., 2008; Daumé III and Jagarlamudi, 2011) and proposed unsupervised statistical machine trans- lation (bilingual lexicon is a byproduct) with only monolingual corpora (Ravi and Knight, 2011; Nuhn et al., 2012; Dou and Knight, 2012). In the bilingual lexicon induction (Koehn and Knight, 2002; Haghighi et al., 2008; Daumé III and Jagarlamudi, 2011), with the help of the or- thographic and context features, researchers adopted an unsupervised method, such as canon- ical correlation analysis (CCA) model, to auto- matically induce the word translation pairs be- tween two languages from non-parallel data only requiring that the monolingual data in each lan- guage are from a fairly comparable domain. The unsupervised statistical machine transla- tion method (Ravi and Knight, 2011; Nuhn et al., 2012; Dou and Knight, 2012) viewed the trans- lation task as a decipherment problem and de- …

Content truncated to 500 words