A Syllable-based Name Transliteration System

Abstract

This paper describes the name entity transliteration system which we conducted for the "NEWS2009 Machine Transliteration Shared Task" (Li et al 2009). We get the transliteration in Chinese from an English name with three steps. We syllabify the English name into a sequence of syllables by some rules, and generate the most probable Pinyin sequence with the mapping model of English syllables to Pinyin (EP model), then we convert the Pinyin sequence into a Chinese character sequence with the mapping model of Pinyin to characters (PC model). And we get the final Chinese character sequence. Our system achieves an ACC of 0.498 and a Mean F-score of 0.786 in the official evaluation result.