关键词: 自然语言处理, 问答系统, 问句分析, 复杂问句, 事件

Abstract:

The complex questions in the question answering ( QA) system usually involve much more event information. Proper processing of the event content is important for the accuracy improvement of QA system. In this paper,an analysis method of complex questions integrating event information is proposed,in which an event is considered as a complex data object containing multiple arguments,the semantic model of the event is defined,and the
similarity calculation method of the events is obtained. In the investigation,by using event extraction methods,the event information of a complex question is obtained to form event semantic model instances,and the whole semantic information of a complex question is then represented by a vector consisting of several semantic model instances. Consequently,the similarity of questions is calculated according to the vectors of semantic model instances of the event. Experimental results indicate that the semantic information of an event is helpful to the question analysis,and that the proposed analysis method of complex questions integrating event information improves the overall performance of QA system,with the accuracy being up to 78. 6%.

Key words: natural language processing, question answering system, question analysis, complex question, event