Modeling Term Translation for Document-informed Machine Translation

Proceedings of the 2014 Conference on Empirical Methods in Natural Language Processing (EMNLP), pages 546–556, October 25-29, 2014, Doha, Qatar. c⃝2014 Association for Computational Linguistics Modeling Term Translation for Document-informed Machine Translation Fandong Meng1, 2 Deyi Xiong3 Wenbin Jiang1, 2 Qun Liu4, 1 1Key Laboratory of Intelligent Information Processing Institute of Computing Technology, Chinese Academy of Sciences 2University of Chinese Academy of Sciences {mengfandong,jiangwenbin,liuqun}@ict.ac.cn 3School of Computer Science and Technology, Soochow University dyxiong@suda.edu.cn 4Centre for Next Generation Localisation, Dublin City University Abstract Term translation is of great importance for statistical machine translation (SMT), es- pecially document-informed SMT. In this paper, we investigate three issues of term translation in the context of document- informed SMT and propose three cor- responding models: (a) a term trans- lation disambiguation model which se- lects desirable translations for terms in the source language with domain information, (b) a term translation consistency model that encourages consistent translations for terms with a high strength of translation consistency throughout a document, and (c) a term bracketing model that rewards translation hypotheses where bracketable source terms are translated as a whole unit. We integrate the three models into hierarchical phrase-based SMT and eval- uate their effectiveness on NIST Chinese- English translation tasks with large-scale training data. Experiment results show that all three models can achieve sig- nificant improvements over the baseline. Additionally, we can obtain a further improvement when combining the three models. 1 Introduction A term is a linguistic expression that is used as the designation of a defined concept in a language (ISO 1087). As terms convey concepts of a text, term translation becomes crucial when the text is translated from its original language to another language. The translations of terms are often af- fected by the domain in which terms are used and the context that surrounds terms (Vasconcellos et al., 2001). In this paper, we study domain-specific and context-sensitive term translation for SMT. In order to achieve this goal, we focus on three issues of term translation: 1) translation ambigu- ity, 2) translation consistency and 3) bracketing. First, term translation ambiguity is related to trans- lations of the same term in different domains. A source language term may have different transla- tions when it occurs in different domains. Second, translation consistency is about consistent trans- lations for terms that occur in the same document. Usually, it is undesirable to translate the same term in different ways as it occurs in different parts of a document. Finally, bracketing concerns whether a multi-word term is bracketable during transla- tion. Normally, a multi-word term is translated as a whole unit into a contiguous target string. We study these three issues in the context of document-informed SMT. We use document- informed information to disambiguate term trans- lations in different documents and maintain con- sistent translations for terms that occur in the same document. We propose three different models for term translation that attempt to address the three issues mentioned above. In particular, • Term Translation Disambiguation Model: In this …

Content truncated to 500 words