Research On Key Technology In Personalized Search Engine

With the rapid growth of Web information, search engines have become the main tools of information retrieval. But general search engines have the shortcomings of providing same results to different users'retrieval requisition, failing to reflect users'personalized demand. At the same time, the major problems existed in current information retrieval (IR) are that the input queries are usually short and ambiguous which largely affects the performance of IR systems. Therefore, if the search engine can automatically provide a list of related queries for a giving query will help users to amend queries and retrieves the information needed. In this paper, we start with the key technology of personalized search engine, and give a detailed discussion of the individual query recommendation technology; the system provides query recommendation for a giving query. Simultaneously, it will achieve the goal of providing personalized query suggestion by capturing the different user preferences.The technology of information retrieval and personalized search engine are studied exploratory in this paper, and have carried on the simulation experiment. The study mainly includes the following aspects:Firstly, this paper presents a personalized clustering algorithm based on concept extraction by capturing the preferences from different users, which can achieve the goal of providing personalized query suggestion. The experimental results show that the method obtains high precision, can provide the potential resources that the user is interested in, and it can significantly improve the quality of personalized retrieval system.Secondly, by analyzing the URLs that the user clicks, we find that some of the tokens appear in a URL are meaningful, especially those high-quality pages. Therefore, these tokens can give a brief description on the summary or subject of the URL. This paper proposes a personalized query recommendation method based on TF-IQF model and graph clustering, and we do some experiments based on this model.Finally, we analysis the user queries in-depth, and give three representations of a query: (1) clicked documents; (2) associated query; (3) reverse query. Each representation corresponds to a semantic similarity calculation, by setting different methods in different weights; we can get a new meaning of query similarity, in this way, user-query semantic similarity can be calculated using search click information, then a new personalized user-query semantic clustering approach can be used based on user preferences which can achieve the goal of providing personalized query suggestion.