User Model Based On Mutual Information And Bayesian Network And Personal Retrieval

As the popularity of Web application increases,the amount of information on the web grows dramatically and the information on the web is updated frequently.The problem users are faced with today is no longer the lack of useful information,but that of finding information pertinent to their personal needs.Traditionally general Information Retrieval(IR)technologies have satisfied people to some extent,but the all-purpose service can't satisfy any query from different background with different intention at different time,which is still suffering from low recall and precision problem. So personal IR is becoming hotspot in the field of IR.Personal IR is an information serving manner which can satisfy user individual requirement. Through studying users' background and behavior,it returns different information to different user. That is to say,based on the simulation of the user by user model,it provide the proper information. At present,there has been much research on personal IR,but it is not a piece of cake.So far,no mature personal IR system is in use.User model plays a key role in personal IR.After studying the research on persona information serving and user modeling in at home and abroad,we find that the theory of user modeling includes statistics-based and semantic meaning-based in personal information service.The user model based on probability statistics and inferring obey the process of finding user's interests,but with the development of semantic web,it shows shortcomings in the expression of semantic meaning.The user model based on semantic meaning upgrade user's interests from the level of keywords to the level of knowledge.Keywords are linked by semantic meaning,which maked user's interests describe more deeply,but ignore their probability distributions.So,the both of them are one-sided and not exact in depicting user's interests.Utilizing the characteristic of complementary of advantages and disadvantages between the user model based on probability statistics and the user model based on semantic meaning,we introduce semantic mutual information in the user model based on Bayesian Network and then put forward User Model Based on Mutual Information and Bayesian Network applied to personal IR——BMB_PRUM.On the basis of introduction of the related theory,we do the following research. Firstly,the struture and formal description of BMB_PRUM are given.Secondly,we show the process of constructing BMB_PRUM in detail,including the initializing algrithom and extracting user's interests and the updating of BMB_PRUM dynamicly.Thirdly,we intrduce how to use BMB_PRUM to realize personal IR.At last,we verify the effectiveness of BMB_PRUM by a simulated experiment.The research on BMB_PRUM is a beneficial exploration in realizing personal IR. BMB_PRUM has a positive effect on recall and precision.