The Research And Application Of Search Engine Personalization Query Expansion Technology

With the rapid development of open source software (OSS), software development has entered the open-source era. The open source community is used to provide a platform for OSS development and idea-exchange for OSS developers and users. Since there is more and more information in the platform, it is essential to build the effective website search service which helps improve the information-acquire efficiency, resources utilization and personal usage. However, the present website search service adopts the method that merely focuses on keywords; the search result is unsatisfactory and low-precision-rate. To address the problem, this paper designs a new method that features the expansion of query words and the personalized ranking of the search results, thereby improving the precision ratio of the web search.This paper first introduces the search engine technology and personalized search technology, and gives an in-depth analysis of query expansion technique and search results ranking method. Later we design the PQES system, which features the personalized query expansion technique based on Web2.0 community. The system constructs a new layer of personalized query expansion on the basis of the existing website search engine. The layer includes a User Feature Library Module, a Query Expansion Module, and a Personalized Ranking Module. In User Characteristic Library Module, we create a user feature database using the combined technology of Search Log Analysis and Data Mining. In Query Expansion Module, we propose the user-feature based Average Mutual Information (AMI) method which is on the basis of the improved query expansion method of Local Context Analysis. In Personalized Ranking Module, we propose a user search ranking algorithms that integrate the query feature of the users with reference to the Relevance Score Fusion technology in the search results ranking of the meta-search engine.In the user characteristic timeliness based query expansion method, we propose the UCTQE algorithm which takes into account of the user characteristic timeliness. The experiments show that algorithm is in average precision ratio of the search results 17.148% compared to the Local Context Analysis. In Personalized Ranking Module, we propose the QEC algorithm which is a ranking method that computes the Relevance Score Fusion with consideration to user query characteristic. The experiments show that average precision ratio of the front-page search results of QEC reach 65%, which gives 6.56% boost over the ranking method that features the search results combination used in Profusion meta search in average R-precision.Finally, the article implemented PQES prototype system in the web2.0 open source community platform——Trac, and evaluated the performance of system and verified its effectiveness and availability. The work and contribution of this paper are parts of"863"project"the key technology and system of open source software IP resources library", and have useful theoretical and practical value.