Abstract: In connection with the characteristics of Chinese question answering systems,a method of query expansion is proposed.It is based on the related words for specific question types that are retrieved by pairing statistically the question-answering classes.A method of similarity between question and answering documents based on minimal matching span is implemented.The term frequency and position of information of query and query expansion in the document are fully considered in the computation.The experimental results show that in answering document retrieval of Chinese question answering system,the performance of expanding the query based on related words for specific question types is much better than in unexpanded query.