An Alignment Based Technique for Text Translation between Traditional Chinese and Simplified Chinese

An Alignment based technique for Text Translation between Traditional Chinese and Simplified Chinese Sue J. Ker Department of Computer and Information Science Soochow University Taipei, Taiwan, R.O.C. ksj@cis.scu.edu.tw Chun-Hsien Lin . Department of Computer and Information Science Soochow University Taipei, Taiwan, R.O.C. jslin@seed.net.tw Abstract Aligned parallel corpora have proved very useful in many natural language processing tasks, including statistical machine translation and word sense disambiguation. In this paper, we describe an alignment technique for extracting transfer mapping from the parallel corpus. During building our system and data collection, we observe that there are three types of translation approaches can be used. We especially focuses on Traditional Chinese and Simplified Chinese text lexical translation and a method for extracting transfer mappings for machine translation. 1� Introduction Aligned parallel corpora have proved very useful in many tasks, including statistical machine translation (Brown et al., 1993; Wu and Ng, 1995, Chen et al., 1997; Moore, 2001) and word sense disambiguation (Chang et al., 1996; Chen and Chang, 1998). Traditional and Simplified Chinese are two Chinese writing systems that used by Chinese-speaking communities. Since their typefaces are different, foreigners always view these two languages as a family of cognate languages. Aside from differences in typeface, their encoding schemas are also different. For the conversion of text, special utilities or tools are required for mapping the correspondence between the two schemas. At present, the methods used to undertake this mapping are far from perfect. In general, a table-conversion method is used to translate between Traditional and Simplified Chinese text. There are several problems with this method. First, correspondences between Big5 (Traditional Chinese) and GB2312 (Simplified Chinese) code schemas are not one-to-one. Thus, this method can cause mismatches in character translation. Second, the unit of processing should be words instead of morphemes since the meaning of a morpheme can be very ambiguous. Third, conventional language usage is also quite different between Traditional and Simplified Chinese. To tackle these difficulties, it thus seems wise to acquire a set of word mappings between Traditional Chinese and Simplified Chinese from parallel corpora automatically. This paper discusses the issues mentioned above, and especially focuses on Traditional and Simplified Chinese text lexical translation and a method for building a synonym thesaurus meaning-translation. The character set standards used in this paper are Big5 code for Traditional Chinese and GB2312 code for Simplified Chinese; nevertheless, these issues are code-independent. Simply categorized, there are three methods for Traditional Chinese and Simplified Chinese conversion and translation, each of which satisfies different purposes. These include code schema conversion, word translation, and semantic translation, all of which are described in the present paper. 2� Motivation As the Internet and World Wide Web become increasingly popular, text documents in electronic form are becoming ever more abundant. Therefore, there is a rapidly developing demand for the translation of text between Traditional Chinese and Simplified Chinese. In addition, there are vast numbers of Traditional Chinese and Simplified Chinese texts in electronic form that require an automatic 147 conversion system for …

Content truncated to 500 words