问答系统中意见型疑问句的分类方法研究

A study of opinion question sentence classification in Question & Answering system

问答系统是自然语言处理领域中一个非常热门的研究方向,当前主流的问答系统主要处理的问题类型大多是问时间、地点、人物等事实性疑问句。本文针对用户提出的对特定领域的产品好坏发表疑问情感性问题作出处理,结合模式匹配算法拥有较高的准确性和语言模型算法拥有较高的召回率,创新性地提出意见型疑问句的分类方法:模式匹配算法结合语言模型算法生成混合模型。首先利用模式匹配算法对问句进行分类,对于不能处理的问句,交给一元或二元语言语言模型算法进行分类。此模型通过试验证明了其有效性。

Question ＆amp; Answering is a very hot area of natural language processing. The current mainstream of Question ＆amp; Answering systems are based on time,place,character. In this paper,we dispose the emotional issues which proposed by the users in a special domain. In order to get a high precision and recall,we combined pattern matching and some language model algorithms,and put forward a classification of opinion-question method. First,we use pattern matching to make a classification of the question,if it doesn＇t work, make another classification by the unigram and bigram language model. At last, The experiment proved the validity of this model.