Query Expansion Research Based On Ontology And User Log

With the explosive growth of the Internet, how to obtain the information whichusers really want from the large amount of information has become increasinglydifficult. The search engine to some extent solved the problem of a user to find usefulinformation. However, when using a search engine, the user often can not accuratelyexpress their query intent, and the search engine always get low precision and unableto return useful information because of the query words improper use or query is tooshort. Therefore, extending the user’s query has become very urgent.Query Expansion technology has undergone decades of development, domesticand foreign researchers have proposed a variety of query expansion methods.However, these common methods often can not understand user input from thesemantic level during the expansion, and, because of its origin of the extend word isuncertain, easy to join query-independent word cause query drift problem. In thispaper, a new query expansion method based on ontology and user log is proposed.Expand user queries using domain ontology from the semantic level to the formationof the initial expanded concept of set, combined with user query logs set forsecondary screening of the initial set by use of co-occurrence analysis. The maincontent of this paper are as follows:(1)Described the research background and significance, Analysised researchprogress and shortcomings of the current query expansion technology, introducedrelevant background knowledge and theory of the subject, which lays a theoreticalfoundation for later research work.(2)Proposed a concept semantic similarity formula combined with domainontology. Calculating semantic similarity of the initial expansion terms, which extendthe user’s query from the semantic level.(3)Proposed a user log-based word co-occurr ence formula for the wordco-occurrence calculation of initial expansion terms, and the calculation results asextended term word co-occurrence weights. Finally, combine extended term’ssemantic similarity weights and word co-occurrence weights to secondary screening,in order to avoid the query drift problem which may lead by initial expansion.(4)Implemented a prototype system based on ontology and user log-based queryexpansion algorithm proposed in this paper, combined with domestic hardware and software service tracking system. Made a brief presentation of the overall frameworkof the prototype system and individual modules, discussed the core data structures andalgorithms of each module in detail. Finally, the prototype system were comparedwith experimental test. Experimental results show that compared with the traditionalquery expansion method, This method not only ensures good robustness but alsoeffectively improve the retrieval accuracy.