基于最大熵的汉语人名地名识别方法研究  

Research on Chinese Person Name and Location Name Recognition Based on Maximum Entropy Model

作　　者：钱晶[1] 张杰 张涛[2] 

机构地区：[1]复旦大学计算机科学与工程系,上海200433 [2]上海财经大学信息管理与工程学院,上海200933

出　　处：《小型微型计算机系统》2006年第9期1761-1765,共5页Journal of Chinese Computer Systems

基　　金：国家自然科学基金项目(60203010)资助.

摘　　要：构建了一个基于最大熵原理的汉语人名地名自动识别混合模型.该模型分为训练和识别两个模块.先从训练语料中抽取特征,利用最大熵方法对特征进行训练.然后使用经过训练的特征,并结合动态词表和少量规则,对测试文本中的汉语人名地名进行识别.达到了比较满意的识别效果.最后对实验结果进行了分析.This paper constructs a hybrid model for Chinese person name and location name automatic recognition, which is based on the maximum entropy principle. The model consists of a training module and a recognizing module. At first, features are extracted from the training corpus. The maximum entropy principle is employed to train the features. Then the trained features together with a dynamic-word-list and a simple rule-base are used to recognize Chinese person names and location names in the testing corpus. The experimental results are satisfying and have been analyzed at the end of this paper.

关 键 词：最大熵模型 专有名词识别 特征提取 语言学规则 

分 类 号：TP393[自动化与计算机技术—计算机应用技术]