Research On Personality Oriented Web Search

With the swift development of Internet, how to find useful information rapidly in such a large-scaled Internet is becoming more and more important. General search engines somewhat satisfy users'search needs. However, they do not consider users'interests or background. Personalized Search Techniques emerge for this purpose, which can provide personalized search results for different interests and background users.This thesis focuses on the research of personalized web search. Firstly, we analyze related feedback techniques, which is one of the key techniques of personalized search. Then, we construct two mathematics models to describe the relationship between users'browsing behaviors and interests. Due to the diversities of browsing behaviors among different users, this thesis designs an algorithm to construct and maintain the relationship of user's behaviors and interests. This algorithm can consider user's own characteristics and gain user's interesting WebPages by combining implicit methods and explicit methods.Secondly, we present another key technique of personalized search: profile exploitation. We provide a two-level interest model to describe user's interest set, which effectively integrates user's current interests and past interests to process the search results.Based on above analyses and research, this thesis designs and implements a personalized search system. We introduce the key techniques about this system, such as Browser Helper Object, webpage text process, Google's API and so on..In order to investigate the diversities of browsing behaviors among different users, we design a corresponding experiment. The result testifies that there are evident differences of users'browsing time, mouse moving and scrolling. Meanwhile, we also use an experiment to evaluate the performance of two-level interest model.