多特征融合的中文问答系统答案抽取算法  

Answer Extraction Algorithm of Chinese QA System Based on Multi-feature Fusion

作　　者：唐朝霞[1] 

机构地区：[1]淮阴工学院计算机工程学院,江苏淮安223003

出　　处：《贵州大学学报（自然科学版）》2011年第5期80-83,共4页Journal of Guizhou University:Natural Sciences

基　　金：江苏省高校自然科学基金(06KJD520024)

摘　　要：随着互联网的迅速发展和Web2.0概念的提出,问答系统以直接返回给用户精确的答案而逐渐成为一种新的信息检索技术。由于问句都是自然语言的形式,涉及到对问句的语义理解及相似度的判断。本文提出了一种基于问句的表层和语义相似度计算方法,通过聚类去除冗余信息,再通过熵的特征计算权值,最后融合多种特征计算问句相似度,进行答案抽取。实验证明,这种方法能够有效地提高答案抽取的精度和效率。With the rapid development and the appearance of concept of Web2.0, because of the exact answer directly, QA system has become a new information retrieval technology. As the questions are in the form of natural language ,Questions relate to the semantics understanding and similarity judgments. In this paper, based on question＇s surface and semantic similarity calculation method, removal of redundant information was carried out by clustering,the weight of features was calculated by entropy,finally question＇s similarity was calculated by integration of multiple features for answer extraction. Experiments show that this method can effectively improve the answer extraction accuracy and efficiency.

关 键 词：问答系统 问句相似度 聚类 答案抽取 

分 类 号：TP311[自动化与计算机技术—计算机软件与理论]