Research Of Query Expansion Technology Based On Trust Network

While the amount of information on the Internet is increasing rapidly, how to get theinformation which meets the users’ demand effectively from the huge information house isbecoming the focus of attention. In the field of information retrieval, query expansion is oneof the important methods to solve the problem which called "word mismatch" and improvethe retrieval efficiency. It adds high-related words into the original query, in order to reducethe impact on the retrieval performance which is caused by the word ambiguity and the shortlength of quires. At present, most query expansion methods do not distinguish different users.Because of the poor degree of personalization, these methods cannot meet the different users’query requirements. The traditional log-based method finds the words with high degrees ofrelevant to the original query based on a large number of user history queries. The statisticalcalculation for the great amount of users is lack of personalized treatment, and cannot providetargeted query results for different users. In addition, the mining on the log information isinsufficient, so the improvement for the search effectiveness is not obvious.Aiming at the problems mentioned above, a query expansion model based on trustnetwork is proposed in this dissertation to select trusted users as the source users of expansionand mine the searching log deeply to improve the expansion efficiency. Firstly, the authorproposes a trust network based query expansion model by taking advantage of the users’social relationship to make the sources abundant, and improve the accuracy of queryexpansion and the user satisfaction. Secondly, a trust computing method which combinessocial trust with similar trust is proposed to make sure that the trust relationship between usersis based on the theme relevance. Thirdly, by mining the original retrieval results rankings andclick rankings in the searching logs, a query expansion method based on user logs and trust isproposed. At last, experimental results show that the novel query expansion model which isproposed in this dissertation can improve the retrieval performance effectively and the usersatisfaction.