西方姓名译名的自动识别研究  

Research on Automatic Transliterated Person Names Identification

作　　者：孔祥龙[1] 卢鹏[1] 陆汝占[1] 刘绍明[2] 

机构地区：[1]上海交通大学计算机科学与工程系,上海200030 [2]富士施乐公司研究本部

出　　处：《计算机仿真》2004年第8期145-148,共4页Computer Simulation

摘　　要：西方姓名译名的自动识别为汉语自动分词不可或缺的组成部分。该文以西方姓名译名用字信息为基础,充分利用标准词表来限制西方姓名译名的过度生长能力,并使用首尾逼近和局部频率等方法来进一步改善识别效果,并且针对西方姓名译名的特点进行了有针对性的处理。对真实语料进行测试,正确率达到96.87%,召回率达到97.20%。Transliterated person names identification is the necessary part of Chinese word segmentation. Based on the information of character that transliterated person names used, this paper makes good use of the standard lexicon to limit the ability of transliterated names to grow, and uses the method, such as modifying potential names based on the character used by the head and tail of transliterated names, local frequency and so on, to further improve the result, also does some special processing in terms of the unique feature of transliterated person names. After tested on real corpus, the precision rate and recall rate are respectively 96.87% and 97.02%.

关 键 词：西方姓名译名 专名识别 标准词表 局部频率 

分 类 号：TP391.43[自动化与计算机技术—计算机应用技术]