Translating Chinese Unknown Words by Automatically Acquired Templates

International Joint Conference on Natural Language Processing, pages 839–843, Nagoya, Japan, 14-18 October 2013. Translating Chinese Unknown Words by Automatically Acquired Templates Ming-Hong Bai1,2 Yu-Ming Hsieh1,2 Keh-Jiann Chen1 Jason S. Chang2 1 Institute of Information Science, Academia Sinica, Taiwan 2 Department of Computer Science, National Tsing-Hua University, Taiwan mhbai@sinica.edu.tw, morris@iis.sinica.edu.tw, kchen@iis.sinica.edu.tw, jason.jschang@gmail.com Abstract In this paper, we present a translation template model to translate Chinese unknown words. The model exploits translation templates, which are extracted automatically from a word-aligned parallel corpus, to translate unknown words. The translation templates are designed in accordance with the structure of unknown words. When an unknown word is detected during translation, the model applies translation templates to the word to get a set of matched templates, and then translates the word into a set of suggested translations. Our experiment results demonstrate that the translations suggested by the unknown word translation template model significantly improve the performance of the Moses machine translation system. 1 Introduction Automatic translation of unknown words is still an open problem. As a result, most statistical machine translation (SMT) systems treat such words as unknown tokens and leave them untranslated. (Koehn et al., 2003; Chiang, 2005; Koehn et al., 2007) The unknown word translation problem has generated considerable interest in recent years. Some works (e.g., Callison-Burch et al., 2006; Marton et al., 2009; Mirkin et al., 2009) focus on finding in-vocabulary paraphrases, which are then used as bridges to translate target unknown words. Li and Yarowsky (2008) proposed an unsupervised method for extracting the mappings from Chinese abbreviations and their full-forms. The method exploits the full-forms as bridges to translate the abbreviations. A prerequisite of the above methods is that the unknown words must have paraphrases (or full-forms). However, many types of unknown words do not have paraphrases (full-forms) naturally. In contrast to paraphrasing methods, Huang et al. (2011) developed a sublexical translation method that translates an unknown word by combining the translations of its sublexicals. However, to deal with the reordering problem, the model combines the translations of sublexicals by considering both straight and inverse directions and uses a language model to select the better one. The ordering is generally morphological structure dependent, but language models only select the most fluent order without considering morphological constraints. In this paper, we propose a translation template model to translate Chinese unknown words. Our model has a number of advantages. First, the translation templates can be extracted automatically from a word-aligned parallel corpus. Second, the word order information is encoded in the templates, so we can compose the translation of an unknown word in a more reliable order. Finally, the expansion of the non- terminal symbol in the translation templates is flexible. The remainder of this paper is organized as follows. In the next section, we introduce the proposed translation template model. In Section 3, we describe the experimental setup; and in Section 4, we evaluate the translations of unknown words derived by our model. Section 5 contains some concluding remarks. 2 Translation Template …

Content truncated to 500 words