一种基于位置概率模型的中文人名识别方法  

A METHOD OF CHINESE PERSON NAME RECOGNITION BASED ON POSITION PROBABILITY MODEL

作　　者：孟伟涛[1] 张蕾[1] 张晓孪[1] 李海军[1] 

机构地区：[1]西北大学信息科学与技术学院,陕西西安710127

出　　处：《计算机应用与软件》2008年第4期187-189,共3页Computer Applications and Software

摘　　要：提出了一种基于位置概率模型的中文人名识别算法。系统的知识源来自于两个方面:人名列表以及标注语料库中提取的人名的左右边界词语。识别过程是:首先根据位置概率模型识别出篇章中可能的人名,然后扩散到整个篇章来召回遗漏人名,最后附加几条启发式规则来对结果进行修正。对40篇新闻语料共计120KB进行开放测试,准确率达80.5%,召回率为76.1%。An effective arithmetic based on position probability model for recognizing Chinese person names is proposed. The knowledge source of the system comes from two aspects, the person name list and the boundary words of person names that we extracted from tagged corpus. The recognition process is as follows：Firstly, the possible person names from the passage are recognized by position probability model. Then, the recognized names are used to recall the omitted names of the passage. Finally, a few rules are appended to modify the recognition results. The method is tested on 40 pieces of news articles（ with 120KB data）. The precision of the test is about 80.5% ,and the rate of recall is around 76.1%.

关 键 词：命名实体识别 人名识别 位置概率模型 词法分析 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术] TD922.7[自动化与计算机技术—计算机科学与技术]