摘要
问答系统特别是近年来流行的社区问答系统是信息检索与自然语言领域研究的热点。问题与答案句匹配及相似问题检索是中文问答系统研究的关键问题之一，其本质是基于相似度计算方法，从已解决的历史问答对中找出与查询问题相关的问题。本文以含有事件（动作）信息的复杂类问句为研究对象，提出了一种基于问题语义表征的问题相似度计算方法，问题语义表征结构由疑问焦点块、问题主题块和问题事件块组成，将一个查询问题与历史问题之间的相似度分解为三个主要语义组块之间相似度。实验结果表明，该方法具有一定的有效性。
Question Answering, especially community Question Answering, is a research hotspot in information retrieval and natural language processing in recent years. The matching between question and answer, similar question retrieval are the key problems of Chinese Question answering. The task of similar question retrieval is to compute the similarity of the queried question and the historical questions which have been solved by other users. In this paper, a kind of questions with event information is emphasized and the method of similarity computation based on question semantic representation model is offered. The structure of question semantic representation is made up of question focus chunk, question topic chunk and question event chunk. So the similarity computation between the query and historical questions is divided into similarity of three chunks. In the end, we apply this method to the experiment of similar question retrieval and the results show this method is efficient.
出处
《情报学报》
CSSCI
北大核心
2014年第10期1099-1107,共9页
Journal of the China Society for Scientific and Technical Information
基金
国家自然科学基金资助项目（61371194）
北京市优秀人才培养资助项目（2013D005017000006）.
关键词
问题语义表征
相似度计算
语义组块
相似问题检索
question semantic representation
similarity computation
semantic chunk
similar question retrieval
