Research Of An Information Retrieval Algorithm Based On The Relevance Of Mobile Search Users

With the development of wireless technology and mobile computing power, users of mobile search service are in the rapidly growing population. Confined by the wireless bandwidth, mobile search require higher precision of the return information, comparing with the Web search while users are also increasingly urgent for personalized search service. How to improve the mobile search queries on the accuracy of the return information to meet the needs of search users becomes an item urgently needed to be solved.After systematically reviewing information retrieval algorithm, the Internet information classification technology and mobile search user behavior mode, this paper synthetically introduces the research of an information retrieval algorithm based on the relevance of mobile search users. Based on the relationship between mobile search users and their search habits, the algorithm makes use of cross-matching approach to the screening of compound corpus information to improve the recall of information to achieve the purpose of personalized search.The main contributions and innovations of this paper are as follows:In the first place, the paper introduces the basic information of retrieval models and the traditional algorithms, including the text-based method, the Link-based method and the amalgamation-based method, proposes the merits and defects of the algorithms and demonstrates that the amalgamation-based method is the way to improve search precision. Some related technologies in Web search have been studied, including focused crawl Nutch, ontology model, intelligent agent technology and mobile search system model.Secondly, the paper established a correlation model of mobile users and directed weighted graph model of the users based on the Graph theory. The indexing effect of User relevance to the information retrieval is also raised. The information retrieval algorithm based on the relevance of mobile search users is designed. The algorithm sorts the documents in the document pool after SVD dimension reduction methods and avoids topic drift phenomenon according to the nature of Markov chains. To improve the sorting accuracy, the duplex filter using the characteristics of the user association is the final step of the algorithm. Based on programming of Java, this paper realized focused crawl Nutch engine and constituted a corpus of data collected on the simulation algorithm. According to mandate of the Conference TREC, we used HITS algorithm for comparison. The algorithm proved its efficiency and feasibility by experiment results.