Sistema de recomendação TV

Domains like the television suffer changes every day, it becomes difficult for a user to be up to date with everything that is happening. The trend most of the times is ignoring that growth and simply watch what’s familiar, the same channels and the same programs. In the worst cases, some turn solely to Web domains in order to download the programs they like. In this document will be presented an implementation of a television recommender based on information retrieved from the Meo STB (Set-top-box). Recommendation is technically information filtering, content-based in this case, whose main goal is helping the users to discover new programs they might like. The information extracted from the STB allows the detection of the preferences of the users and build upon them a profile. Knowing their tastes it is then possible to infere and recommend all the products that the television has to offer. The work developed covers only programs, infering about their type, genres and even channels. The watched content is extracted from the STB in XML format. This content contains information about the time and date of the visualization, title, synopsis, channel and total duration. If a visualization is detected, the algorithm will classify it according to certain genres which depend on the type of the program and its synopsis, in case that fails and if the type of the program is film or serie the algorithm will then try to extract the information required to the Website IMDB. On the other hand, there’s the information available in the EPG which include more information about the list of the candidates to the classification. This list of candidates is classified the same way as the visualizations. Having the list of candidates it is then possible to filter it according to it’s content. The result is a final list for recommendation. The work developed on this dissertation is only the beginning for a more complex recommendation system which most likely will integrate collaborative filtering. This work can be useful also for possible extensions beyond normal programs, like targeted publicity or to reinforce a pay-per-view service