Abstract: Answer extracting is the key part of question-answering system. The basic structure and realization of Chinese question-answering system are introduced. Based on the statistic feature of keyword frequency, the distribution of keywords in question and sentence is considered. And a similarity computation method between question and sentence, which combines vector space model (VSM) and keyword minimal matching span is proposed. According to question type and the similarity calculated above, answer-extracting experiment for Chinese factoid question is done. The experiment result shows that the method presented in this paper gets a very good effect

Key words: Question-answering system; Answer extracting; Similarity; Vector space model; Minimal matching span

摘要： 答案提取是问答系统的关键部分，文章介绍了汉语问答系统的基本结构及其实现过程，以问题和答案中关键词的词频统计特性为基础，进一步考虑问题和句子中关键词位置分布信息，提出了一种结合向量空间模型(VSM)和关键词最小匹配距离的问题和句子相似度的计算方法。并以相似度为基础，结合问题类别，对汉语基于事实的简单陈述问题进行了答案句子提取实验，结果表明该方法有较好的效果。

关键词: 问答系统；答案提取；相似度；向量空间模型；最小匹配距离