基于问句表征的web智能问答系统  

Question-answering system based on question token

作　　者：康海燕[1] 李飞娟[2] 苏文杰[2] 

机构地区：[1]北京信息科技大学信息管理学院,北京100192 [2]北京信息科技大学计算机学院,北京100192

出　　处：《北京信息科技大学学报（自然科学版）》2011年第1期36-41,共6页Journal of Beijing Information Science and Technology University

基　　金：国家自然科学基金面上项目(60873013);北京市高校学术创新团队建设计划项目(PHR201106133)

摘　　要：问答系统现已成为自然语言处理领域和信息检索领域的一个重要分支,问句表征是问答系统的重要组成部分。针对问句分析的不足,采用改进的贝叶斯和疑问词相结合的方法,对问句进行分类,提取问句的类型;在初次查询的基础上,根据问句的类别进行关键词扩展,同时,加入新词识别的功能,即提出一种问句的表征方法,对问句进行充分、全面地理解;基于以上算法,在J2EE环境下开发出一个问句表征的web智能问答系统。实验结果表明,文中的问句表征方法能有效改善系统的性能。Question-answering system is an important offset in natural language processing and information retrieval fields. Question token is a key part of this system. First, the united method of Bayes and question words is adopted to classify the question words according to question types. Second, on the basis of initial search, the question words are extended by the question types. At the same time, the function of discriminating new word is employed, i. e. an improved method of question expression is pro- posed, which can generally understand the question. Finally, a question-answering system is put into effect on J2EE based on the above theory and algorithm. Experiment proves that it considerably enhances the intelligence degree of question-answering

关 键 词：问句表征 问答系统 自然语言处理 

分 类 号：TP312[自动化与计算机技术—计算机软件与理论]