Improved Statistical Machine Translation for Resource-Poor Languages Using Related Resource-Rich Languages

Proceedings of the 2009 Conference on Empirical Methods in Natural Language Processing, pages 1358–1367, Singapore, 6-7 August 2009. c⃝2009 ACL and AFNLP Improved Statistical Machine Translation for Resource-Poor Languages Using Related Resource-Rich Languages Preslav Nakov Department of Computer Science National University of Singapore 13 Computing Drive Singapore 117417 nakov@comp.nus.edu.sg Hwee Tou Ng Department of Computer Science National University of Singapore 13 Computing Drive Singapore 117417 nght@comp.nus.edu.sg Abstract We propose a novel language-independent approach for improving statistical ma- chine translation for resource-poor lan- guages by exploiting their similarity to resource-rich ones. More precisely, we improve the translation from a resource- poor source language X1 into a resource- rich language Y given a bi-text contain- ing a limited number of parallel sentences for X1-Y and a larger bi-text for X2-Y for some resource-rich language X2 that is closely related to X1. The evaluation for Indonesian→English (using Malay) and Spanish→English (using Portuguese and pretending Spanish is resource-poor) shows an absolute gain of up to 1.35 and 3.37 Bleu points, respectively, which is an improvement over the rivaling approaches, while using much less additional data. 1 Introduction Recent developments in statistical machine trans- lation (SMT), e.g., the availability of efficient im- plementations of integrated open-source toolkits like Moses (Koehn et al., 2007), have made it pos- sible to build a prototype system with decent trans- lation quality for any language pair in a few days or even hours. In theory. In practice, doing so requires having a large set of parallel sentence- aligned bi-lingual texts (a bi-text) for that lan- guage pair, which is often unavailable. Large high- quality bi-texts are rare; except for Arabic, Chi- nese, and some official languages of the European Union (EU), most of the 6,500+ world languages remain resource-poor from an SMT viewpoint. While manually creating a small bi-text could be relatively easy, building a large one is hard, e.g., because of copyright. Most bi-texts for SMT come from parliament debates and legislation of multi-lingual countries (e.g., French-English from Canada, and Chinese-English from Hong Kong), or from international organizations like the United Nations and the European Union. For exam- ple, the Europarl corpus of parliament proceed- ings consists of about 1.3M parallel sentences (up to 44M words) per language for 11 languages (Koehn, 2005), and the JRC-Acquis corpus pro- vides a comparable amount of European legisla- tion in 22 languages (Steinberger et al., 2006). The official languages of the EU are especially lucky in that respect; while this includes such “classic SMT languages” like English and French, and some important international ones like Span- ish and Portuguese, most of the rest have a limited number of speakers and were resource-poor until recently; this is changing quickly because of the increasing volume of EU parliament debates and the ever-growing European legislation. Thus, be- coming an official language of the EU has turned out to be an easy recipe for getting resource-rich in bi-texts quickly. Of course, not all languages are that ‘lucky’, but many can still benefit. In this paper, …

Content truncated to 500 words