English-Chinese Name Transliteration with Bi-Directional Syllable-Based Maximum Matching

English-Chinese Name Transliteration with Bi-Directional Syllable-Based Maximum Matching  Oi Yee Kwong Department of Chinese, Translation and Linguistics, City University of Hong Kong, Tat Chee Avenue, Kowloon, Hong Kong Olivia.Kwong@cityu.edu.hk Abstract. In this paper, we propose a simple and intuitive yet linguistically and practically motivated method for English-Chinese name transliteration generation. Our system is essentially a syllable-based Maximum Matching system. It uses the Onset First Principle to syllabify English names and align them with Chinese names. The bilingual lexicon containing aligned segments of various syllable lengths subsequently allows direct transliteration by chunks. The proposed method was tested on the data from the shared task of the Named Entities Workshop 2009. The results suggest that Forward Maximum Matching performed slightly better than Backward Maximum Matching, but when used together much better results comparable to those of state-of-the-art methods could be attained. Keywords: Name transliteration, Syllable-based Maximum Matching.  The work described in this paper was supported by a grant from City University of Hong Kong (Project No. 7008004). Copyright 2011 by Oi Yee Kwong 1 Introduction In this paper, we propose a syllable-based Maximum Matching method for English-Chinese name transliteration generation. Compared to most other popular methods for the purpose, our approach is relatively simple and intuitively appealing, but at the same time it has strong linguistic and practical motivation. Our system is simply trained on a set of English-Chinese name pairs. Two major principles underlie its design. On the one hand, syllables form the primary units in transliteration. While the rendition of a name in the target language is expected to phonemically resemble its pronunciation in the source language, difference in phonotactics between the two languages would be compromised during transliteration, and incompatible phonological structures might result in extra syllables in the transliterated name. On the other hand, name transliteration tends to be more conventional than innovative, such that the same chunk is often transliterated in the same way even when embedded in different names. Thus, with our proposed method, the Onset First Principle in phonology was used to syllabify English names and align them with the Chinese renditions. A bilingual lexicon containing segment pairs of various syllable lengths was then produced from the aligned names. This lexicon was subsequently used in transliteration, during which a source name was first syllabified and then segmented using Maximum Matching with syllables as the basic units. Target candidates were generated by looking up the bilingual lexicon. Both directions of the Maximum Matching and various ways of ranking the candidates have been tested. Experimental results show that while Forward Maximum Matching performed slightly better than Backward Maximum Matching, their 25th Pacific Asia Conference on Language, Information and Computation, pages 11–19 11 combined use could lead to much better results comparable to those produced by state-of-the- art methods. We will briefly review related work in Section 2, and introduce the datasets used in this study in Section 3. The system will be described in Section 4. Experiments and results will be reported in Section 5, …

Content truncated to 500 words