Research On Cross-application User Modeling Based On Linked Data

Social network is becoming more and more popular in modern world, which gradually become our necessary instrument. Every user always registers in many social network platforms, so that there is a large amount of individual information which is left in these social network platforms. How to fully mine a large number of user data and apply it to personalized service is a very hot topic in scholar circles.This paper propose a user modeling technology is a key research point which could resolve above problem. However the user modeling cross application provide a more completed user view and more value-added information of user for all relevant application. Furthermore, under multiple user environment, user modeling also need more completed user model view and value-added information.This paper firstly put forward to an extensible core user model(CUM+E) which could depict the collective property of 2 applications and personalized features of user in certain field, and study the connection among pieces of information or cases of user who registered in different application circumstance, release the cases with the form of linked data to realize the sharing and reuse of data, and construct a virtual global user model with this method according to the Linked Data’s advantage. It’s potentially a huge convenience for us when searching or querying any user’s information.Next we research a relationship between a model fragment of users who register in many different applications and the linked data which is released with these fragments. We propose a kind of linked data driven method which could integrate usermodel from different data resource, and determine an effective modeling process which is assisted by a framework. In simple terms, we propose a user modeling means which is driven by linked data to construct connection among user model fragments of one user, and this user has registered in different applications.At last we propose a method of user recognition, determine the weight of public properties in user archive by introducing entropy value, and then use similarity measurement to obtain the eventual result. Thus it would take the foundation for establishing linked relationship when lacking the unique identifier cross different applications.