基于问句相似度的本体问答系统  

Ontological Question Answering System Based on Sentence Similarity

作　　者：刘汉兴[1] 刘财兴[1] 林旭东[1] 

机构地区：[1]华南农业大学信息学院,广东广州510642

出　　处：《广西师范大学学报（自然科学版）》2010年第1期88-91,共4页Journal of Guangxi Normal University:Natural Science Edition

基　　金：国家863计划资助项目(2006AA10Z246)

摘　　要：本体问答系统需要实现从自然语言问句到本体查询语句的转换,目前的解决方法主要有自然语言接口和问句相似度方法。针对现有问句相似度方法在本体问答系统中应用的不足,设计了改进的相似度计算方法。通过建立常问问题的查询模式集合,综合考虑问句的统计、语义、结构特征计算目标问句的相似度,分别以自动选择和用户交互两种方式选择目标问句的查询模式,并将其转换成实际SPARQL查询语句,最终检索本体及抽取出答案。两种方式的准确率分别为83.8%和92.1%。The ontological question answering system needs to convert the questions in natural language into ontology query statements ,and the present solutions are mainly natural language interface and question similarity. To overcome the disadvantage of sentence similarity applied in ontological QA system,the improved method of computing similarity is designed. Via building the set of query pattern for frequently asked questions ,the similarity of target question is computed based on the statistic,semantic and structure of question. The query pattern of target question is chosen in automatic and interactive mode separately,then it is converted into real SPARQL query statements,and finally ontology is retrieved to extract answers. The accuracy rates of the two modes are 83.8% and 92.1% separately.

关 键 词：问句相似度 问答系统 本体 查询模式 

分 类 号：TP182[自动化与计算机技术—控制理论与控制工程]