Name Origin Recognition Using Maximum Entropy Model and Diverse Features

Name Origin Recognition Using Maximum Entropy Model and Diverse Features Min Zhang1, Chengjie Sun2, Haizhou Li1, Aiti Aw1, Chew Lim Tan3, Xiaolong Wang2 1Institute for Infocomm Research, Singapore {mzhang,hli,aaiti} @i2r.a-star.edu.sg 2Harbin Institute of Technology, China {cjsun,wangxl} @insun.hit.edu.cn 3National University of Singapore, Singapore tancl@comp. nus.edu.sg Abstract Name origin recognition is to identify the source language of a personal or location name. Some early work used either rule- based or statistical methods with single knowledge source. In this paper, we cast the name origin recognition as a multi-class classification problem and approach the problem using Maximum Entropy method. In doing so, we investigate the use of differ- ent features, including phonetic rules, n- gram statistics and character position infor- mation for name origin recognition. Ex- periments on a publicly available personal name database show that the proposed ap- proach achieves an overall accuracy of 98.44% for names written in English and 98.10% for names written in Chinese, which are significantly and consistently better than those in reported work. 1 Introduction Many technical terms and proper names, such as personal, location and organization names, are translated from one language into another with approximate phonetic equivalents. The phonetic translation practice is referred to as transliteration; conversely, the process of recovering a word in its native language from a transliteration is called as back-transliteration (Zhang et al, 2004; Knight and Graehl, 1998). For example, English name “Smith” and “史密斯 (Pinyin 1: Shi-Mi-Si)” in 1 Hanyu Pinyin, or Pinyin in short, is the standard romaniza- tion system of Chinese. In this paper, Pinyin is given next to Chinese form a pair of transliteration and back- transliteration. In many natural language process- ing tasks, such as machine translation and cross- lingual information retrieval, automatic name transliteration has become an indispensable com- ponent. Name origin refers to the source language of a name where it originates from. For example, the origin of the English name “Smith” and its Chi- nese transliteration “史密斯 (Shi-Mi-Si)” is Eng- lish, while both “Tokyo” and “东京 (Dong-Jing)” are of Japanese origin. Following are examples of different origins of a collection of English-Chinese transliterations. English: Richard-理查德 (Li-Cha-De) Hackensack-哈肯萨克(Ha-Ken- Sa-Ke) Chinese: Wen JiaBao-温家宝(Wen-Jia- Bao) ShenZhen–深圳(Shen-Zhen) Japanese: Matsumoto-松本 (Song-Ben) Hokkaido-北海道(Bei-Hai-Dao) Korean: Roh MooHyun-卢武铉(Lu-Wu- Xuan) Taejon-大田(Da-Tian) Vietnamese: Phan Van Khai-潘文凯(Pan- Wen-Kai) Hanoi-河内(He-Nei) In the case of machine transliteration, the name origins dictate the way we re-write a foreign word. For example, given a name written in English or Chinese for which we do not have a translation in Chinese characters in round brackets for ease of reading. 56 a English-Chinese dictionary, we first have to de- cide whether the name is of Chinese, Japanese, Korean or some European/English origins. Then we follow the transliteration rules implied by the origin of the source name. Although all English personal names are rendered in 26 letters, they may come from different romanization systems. Each romanization system has its own rewriting rules. English name “Smith” could be directly transliterated into Chinese as “史密斯(Shi-Mi-Si)” since it follows the English phonetic rules, while …

Content truncated to 500 words