基于词共现模型的常问问题集的自动问答系统研究  

Question Answering System Based on Frequently Asked Questions Using Co-Occurrence Word Model

作　　者：钟敏娟[1,2] 万常选[1,2] 刘爱红[1] 

机构地区：[1]江西财经大学信息管理学院,南昌330013 [2]江西财经大学数据与知识工程江西省重点实验室,南昌330013

出　　处：《情报学报》2009年第2期242-247,共6页Journal of the China Society for Scientific and Technical Information

基　　金：江西省教育厅科技项目（赣教计字[2007]271）,国家社会科学基金项目（07BTQ025）;全国教育科学“十一五”规划2008年度教育部青年专项课题（ECA080292）共同资助。

摘　　要：在自动问答系统中引入基于Frequendy asked questions（FAQ）的辅助模块满足常见问题的回答是一种有效的手段,其中关键问题是用户提出的问句与FAQ中问句的相似度比较,找出FAQ中最相似的问句,并返回对应的答案。本文将词共现模型引入到问句的相似度匹配中,利用互信息构造共现词汇,同时,结合相关关键词个数及问句长度等信息计算问句之间的相似度。相关实验结果表明,结合词共现模型的FAQ自动问答系统具有较高的准确率和较快的响应速度。It is an effective method to incorporate a frequently asked questions（FAQ）accessorial module into a QA system. The most important question of the system is how to match the user queries and questions of the FAQ corpus.In this paper,cooccurrence word model is described and applied to the similarity matching bewteen the questions,in which term co-occurrence corpus is firslty constructed by using mutual information,and then similarity value between questions is computed through some factors like the number of the relative words and the length of the question.Experiments show that the system based on FAQ, which combined co-occurrence word model possesses higher precision and faster response speed.

关 键 词：自动问答 FAQ 词共现模型 相似度计算 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程] N94[自动化与计算机技术—控制科学与工程]