Behaviors-based User Profiling and Classification-based Content Rating for Personalized Digital TV

This paper proposes a system embedded within digital TVs that aims at TV program recommendation based on descriptive metadata collected from versatile sources. The proposed system comprises a user profiling subsystem identifying user preferences and a user agent subsystem performing content rating. For intelligent implicit TV profiling, a novel scheme for observable TV user behaviors is developed based on linear regression. Furthermore, a new relation-based similarity measure is suggested to improve categorized TV program rating precision. The experimental results show that the content rating precision is enhanced enough by the proposed schemes.