User Experience for Recommendation System for Smart TV

Abstract

This paper describes how user experience can be used to recommend program to user. Smart televisions are becoming more and more popular and together with TV content provider offer hundreds of programs to watch. One cannot have real time overview and do not want to spend time to switch channels to find something to watch. That is why we think that user can find useful to relay on recommendation and so save his time. Recommendation is based on past user experience. But it is not only statistic of watched history. This paper is dedicated to system design which is able to learn user habits and correlation between programs watched in past, weather, day in week and many more aspects which can affect user's tastes. We will use graphical model to explore inferences between features affecting user decisions. Bayes linear regression is used to train and predict future recommendations. Recommendation system in this paper is content based where program information can be extracted from electronic program guide. As the design is for smart TV, we can use additional information from the internet if necessary.