Query Expansion Based On User Interest And Ontology Term Relations

By analyzing the defect of traditional vector space model in retrieve that only consider the matching of single query and document, not considering the user interest and the semantic relations between query and document, ontology is introduced into information retrieval, which has good concept hierarchy and logicality. Based on user interest and relation of ontology terms, we put forward a query expansion method, which aims at returning different results as per the initial intent although different users input the same query, and we theoretically discussed the superiority and feasibility of the method.Term ontology correlation between terms provides a unified quantitative standard to compute the relation of terms in ontology, which is the core content of the whole method and the basis of choosing extension word. So the paper put forward the extended concepts and calculation formula, including term and term, term and group, group and document etc. It guarantees to calculate similarity between query and document based on ontology term relation.The realization of query expansion method based on user interest and ontology term relation mainly has six steps, including calculation and marked interest weight in terms, judgment of web page domain, proportion of interest, query expansion and so on. Interest weight calculation includes term weights, term frequency and reading time, which is the key to analyzing user interest. Extension word has taken three factors into account as correlation of ontology, interest weight, proportion of interest domain, and it is important to optimize search results.By using the recall rate and precision index in experiments, the performance of new method would be compared with the traditional method respectively in twocases of query words, as single domain and multi-domain. Experimental results show that the personalized query expansion can greatly improve the retrieval performance, which is superior to the traditional methods in both the single and multi-domainquery, particularly when the query words belong to multi-domain the advantages were more obvious.