Analysis Of User's Chunk Query In Information Retrieval

How to retrieve information quickly and correctly from internet is now an important research domain. Nowadays, most of the search engines are based on keyword matching, which matches the strings users inputs with stored documents. Therefore search engines output too many information most of which is useless. Understanding the user's intention according to the user's input in information retrieval is a key point to improve the performance of search engines. User's real intention is a whole concept. Analyzing the user's query is to recover user's original query concept by the discrete input. It will improve the search performance remarkably.In the paper, we analyze the user's chunk query. This kind of query is most frequent in users'query log. Conceptual graph is used to represent query concept. In the graph, a vertex represents the concept and a edge represents the relation between concepts. First, we label out the conceptual frame graphs of some query classes by manual sorting. Second, we do some preprocessing before analyzing query input, including segmentation and name entity recognition, both of which are based on web mining. We classify user's query with the method based on complex named-entity recognition and rules, and get the values of the class's attributes from the user's input to fill the slots of conceptual frame graph. So we can get the conceptual graph of user's query.In the paper, we collect lots of user queries in specific classes. These classes have many instances in the log of search engines. We test our system's performance using these query instances. The experiment shows that our system has good analysis result for the query inputs in some given classes.