An approach for delivering personalized ads in interactive TV customized to both users and advertisers

Speaker: Georgia Kastidou

In this paper, we present a model for delivering personalized ads to users while they are watching TV shows. Our approach is to model user preferences, based on characterizing not only the keywords of primary interest but also the relative weighting of those keywords. We combine the results of two separate agents - TMA (TV Monitoring Agent) tracks the kind of shows being watched by the user, for how long and on what days; IMA (Internet Monitoring Agent) captures the keywords of interest to the user, based on browsing activity. The conclusions reached by these two agents are merged into one representation, compared to a characterization of possible ads to be delivered, adjusted to fit into required timeslots. We consider as well the case of providing ads for an entire household of users, making use of the collection of individual profiles. We discuss how our approach results not only in benefit to users but also for the benefit to advertisers.

This is joint work with Robin Cohen