基于CRF和规则的中文医疗机构名称识别  

CRF AND RULES-BASED RECOGNITION OF MEDICAL INSTITUTIONS NAME IN CHINESE

作　　者：张金龙[1] 王石[2] 钱存发[1] 

机构地区：[1]江苏科技大学计算机科学与工程学院,江苏镇江212003 [2]中国科学院计算技术研究所,北京100190

出　　处：《计算机应用与软件》2014年第3期159-162,198,共5页Computer Applications and Software

基　　金：国家社科基金重点项目(10AYY003)

摘　　要：中文机构名称识别是中文自然语言处理的一项重要的研究课题。以中文医疗机构名识别为目标,从因特网上抓取了30余万条机构名和1 500篇包含医疗机构名的文本语料,对医疗机构名的组成特点和上下文信息进行详细的分析,总结出若干模式、规则和特征,最后结合条件随机场模型,建立了一个统计和规则相结合的中文医疗机构名识别模型。实验结果表明,该模型的识别效果很好。The recognition of organisations name in Chinese is an important research topic in Chinese natural language processing. In this paper, we focus on the recognition of medical institutions name in Chinese, by crawling over 0.3 million agency names and 1 500 corpus containing the name of the medical institution from the Internet, we elaborately analyse the composition characteristics of the medical institutions name and the context information, sum up some m0dalities, rules and features. At last, combining the conditional random field model, we build a recognition model of medical institutions name in Chinese which contains both the statistics and the rules. Experimental results show that the recognition effect of the model is very good.

关 键 词：中文机构名识别 条件随机场 命名实体识别 自然语言处理 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]