Research And Implementation Of User Interest Mining For Personalized Services

Recent years have witnessed the great progress of internet technology and its application.The explosive growth of the cyber information has resulted in information overload,which is a tremendous challenge for both web users and service providers.Meanwhile,a large amount of data has been generated and it is now a hotspot to extract valuable rules and knowledge from the data.Personalized service was proposed as a solution to the mismatch problem between informational service and user's demand.Generally speaking,the key to this solution lies in text mining and multi-source data fusion techniques based user interest modeling.However,the current research on user profile remains inadequate in diversified description of user interest and its change.Focusing on user interest modeling,this paper is dedicated to the problem of description and migration of user interest.In this paper,several techniques such as natural language processing,knowledge warehouse,data fusion and text mining are employed to extract interest from user generated content and user behavioral information.At last,a personalized user interest extraction system was designed to validate the feasibility of the proposed method.The main contributions are concluded as follows:(1)Interest description scheme.Due to the fact that the description of the user interest varies from person to person,this paper proposed a detailed interest description scheme based on Open Directory Project.In addition,we proposed an interest mapping method that is capable of mapping diversified user interest into a standardized label space.(2)User Interest modeling.This paper regards user interest as a combination of long and short term interest.Therefore,an interest updating method,consisting of interest extraction and fusion algorithm,is proposed to describe the co-effect of long and short term interest.Moreover,a historical similarity based interest migration detection method is proposed,aiming at describing the time dependent pattern of user's interest with fully consideration of the impact of time factor on interest migration.