Recommender Systems for Interactive TV

Recommender systems alleviate the problem of searching for interesting content in the iTV do- main where users are overwhelmed by the abundance of TV channels, programs and videos. The technologies enabling a personalized viewing experience through TV recommendations are evolv- ing rapidly. In this chapter the newest developments in the domain of recommender systems for personalized TV are presented. The focus is on state-of-the-art technologies that can be used to identify users’ taste by adopting either social-based, collaborative approaches or content-based analysis of TV programs. The chapter also provides a deep understanding of the role played by user-based assessment methods and of the effectiveness of automatic quality metrics (accuracy, errors, coverage, novelty). Finally, a case study is presented describing the integration of a rec- ommender system into the production environment of Fastweb, one of the largest European IP Television providers. The recommender system implements both collaborative and content-based techniques. The algorithms are extensively analyzed by means of off-line and on-line tests, show- ing the effectiveness of the recommender systems: up to 30% of the recommendations to Fastweb TV customers are followed by a purchase, with an estimated lift factor (increase in sales) of 15