基于潜在语义分析的智能答疑系统研究与实现  

Research on Intelligent Question Answering System Based on Latent Semantic Analysis

作　　者：乌庆敏[1] 杨思春[1] 

机构地区：[1]安徽工业大学计算机学院,安徽马鞍山243002

出　　处：《计算机技术与发展》2008年第9期251-252,F0003,共3页Computer Technology and Development

基　　金：安徽省自然科学基金项目(KJ2007B245)

摘　　要：文中研究的是基于常问问题库(FAQ库)的智能答疑系统。FAQ库是很多智能答疑系统中的一个重要组成部分,它把用户常问的问题和相关答案保存起来,对于用户输入的问题,可以首先在FAQ库中查找答案。如果能够找到相似的问题,就可以直接将问题所对应的答案返回给用户。为解决智能答疑系统因词的同义或多义现象而导致的"漏答"或"错答",采用一种基于加权潜在语义分析模型的相似度计算方法。针对特定教育领域的智能答疑系统,改进了反映词与词之间相关性的权值计算。通过对特定课程中常问问题的实验,结果显示明显优于向量空间模型。It is focused on frequency answer questions （short for FAQ） library. FAQ library is one of the key components of QAS. By the means of saving frequency questions and relevant answers, its rate of the processing is very great. While extracting answers in intelligent question answering system, synonymy and polyserny can bring about losing correct answers or extracting wrong answers. In order to solve these problems,adopts a method based on weighted latent semantic analysis model to measure similarity between sentences. Aiming at specific intelligent question answering system, the calculating of weight reflecting relevance between words is improved. With some frequently answering questions,the results show that this method is better than vector space model obviously.

关 键 词：智能答疑 潜在语义分析 相似度 向量空间模型 

分 类 号：TP312[自动化与计算机技术—计算机软件与理论]