A linked data-driven solution to user modeling in a Multi-Application Environment

Abstract

The User Model plays a significant role in many personalized applications. Separated user modeling results in the fragments of the user model being distributed on the Web, but user modeling in a Multi-Application Environment (MAE) is increasingly required to provide a more complete user model view and more value-added user information to all relevant applications. In this paper, a Linked Data-driven user modeling approach is proposed to create links among the user model fragments of the same user who registers in different applications and, from these, construct a virtually global user model. Through these links, user information in one application can be acquired by another. An extendible Core User Model (CUM) is presented and used to describe both the common properties and the domain-specific features of the user. The advantages and disadvantages of this approach are analyzed by comparison with other user modeling methods in MAEs. The proposed solution appears to be more feasible and to offer increased flexibility.