Research Of The Methods Of Domain Knowledge Understanding And Responding

With the popularization of computer network applications, the users'ability of obtaining information from the Internet has been greatly improved. How to get useful knowledge from the millions of various kinds of information sources, understand the domain knowledge and utilize the domain knowledge are becoming research focuses. Question and Answer system which is a form of information retrieval is under development in such circumstances. Compared with the traditional information retrieval form like Google's and Baidu's horizontal search engine,the question answering system which is constructed by facing domain can improve the result's correlation,thus improving the system's retrieval quality.In this paper, we focus on the points of domain knowledge understanding and responsing methods,and start a study on the question answering system,the study is divided into three specific aspects:Fristly, in the aspect of the domain knowledge, we study on the method of agricultural ontology which is constructed from the "thesaurus of agricultural science". Based on this effort, we analyse of the characteristics of question and answer system in the agricultural domain, furthermore, we build an agricultural "question and answer" ontology with the OWL technology.Secondly, in the aspect of the agriculture question understanding,we finish the analysis of the question with the nature language processing and agricultural domain's ontology technology,we obtain the key information about the quetion:question's type, interrogation, keywords, and then,with agricultural ontology we can expand the theme of the quesiton keywords'semantic concept.Thirdly, in the aspect of the agriculture question answering,we use vector space model(VSM)to calculate the similartity of questions. Base on a candidate question set, user's question's feature vector is made of the expansion of the quesiton keywords' concept, the candidate question's feature vector is made of question keyswords, and then, we finish the similarity calculation based on Vector Space Model. By the way, the candidate question set is composed of the the same type of user's question from the "question and answer" agricultural ontology.Base on the research contents above, we buil an agricultural domain question answer prototype system, it includes three modules: Question parsing module, retrieval module, agricultural "question and answer" ontology conversion module. The performance of our system is evaluated by the rate of recall and precision. Following this article's evaluation scheme, the results show that the question answering system's recall rate and precision rate are improved.