Study On User Concept Discovering Method For Exploratory Search

With the rapid development of Web2.0, search engine has been more and more widely used. Search engine could provide qualified search results for queries with clear intention. However, when users lack knowledge of the target area, or the search task itself requires a lot of analysis and summarization, the current search engine will not directly help the user to complete the search process. In this case, users usually need to submit some of the tentative search queries, analyze the returned results, and determine the next step of the search. This searching mode is called exploratory search. For exploratory search, there is no commonly accepted solution. Exploratory searching process could be divided into three stages:search aggregating, discovery supporting, and content synthesizing. Among them, the discovery supporting phase of the main task is to support users find resources that could help them complete their exploratory search process. To support the discovering process, a typical approach is to help users find unknown concepts. Using these concepts, users will be able to find documents related to concepts, and thus complete the exploratory search process.In response to these problems, this thesis bases on the study of faceted search, and proposes the concept of exploratory search concept discovery process. This thesis further studies concept matching methods, concept merging methods, and concept selecting algorithm. For the keywords entered by users, a group of concepts which havemost comprehensive description and are most representative of the target areas are selected to help the user explore the target area.Specifically, based on the study of faceted search, this thesis summarizes the process of exploratory search concepts discovering as:knowledge base building phase, keyword concept sets building phase, concept matching phase, concept merging phase and concept selecting phase. In the stage of knowledge base building, this thesis combines folksonomy and wikipedia to provide knowledge support for concept discovering. In the stage of concept matching, this thesis builds keywords’wikipedia model according to definitions in wikipedia, and match concepts using heuristic rules to obtain concepts matching result set. In the stage of concepts merging, according to the concept matching result sets, this thesis introduces concept merging method base on heuristic. In the stage of concept selecting, according to folksonomy, this thesis builds the <concept resources> relation information network, and proposed concept selecting algorithm based on the RankClus method to cluster concepts and sort nodes. According to the clustering and sorting results, a set of concepts which has most comprehensive description and are most representative of the target area are selected. And as the result of concept discovering, these concepts are returned to users. According to the result of concept discovering, the number of interactions till finding the required documents and the relevance of resulted documents are used to evaluate the original search results and the results yielded by direct sorting method. The number of concepts in concept discovering result set found in user browsed documents are also used to evaluate the proposed method. The experimental results show that the method can efficiently help users explore the target area.