Whether you receive TV programmes through DVB-T, cable or satellite, you may be able to receive hundreds of channels with a wide variety of programs. Traditional program guides, whether printed in magazine form or on-line as a grid based program overviews, do not give you the right level of detail, leading either to endless channel hopping or time lost browsing for interesting programs to watch. We have developed CPTV, a personalized TV scheduler which selects and schedules interesting programs based on a viewers’ preferences.
The program consists of three main components:
- A program selector parses the data feed containing all programmes over the next two weeks to select items which are interesting to the viewer. This is based on a rule set which defines which programs the user likes to watch, and on the programs the user has watched in the past.
- The number of programs that the user might be interested in typically exceeds the capacity of the user to actually watch programs. Also many interesting programs are transmitted several times, either on the same or on alternative channels. Some programs (i.e. sport) must be watched live, as the events are happening. Other programs may be time-shifted on the same day. You may want to watch the evening news a little later on the evening, but clearly not on the next day. Other programs can be recorded and watched at a later time. But for TV series you may want to watch the episodes in the correct order, in order to follow the plot lines. All these scheduling constraint are fed into a personal program scheduler, which knows about recording capabilities and preferences, and which also has access to the users’ calendar to find time periods for TV watching. The scheduler is a constraint based application which produces a recording schedule which may feed directly to your PC or digital video recorder, and a watching schedule, which is transferred to the viewers’ calendar application or mobile phone.
- If the programme selection is purely based on explicit user preferences, you may miss many interesting programs (movies) you have not heard of before. As a third component of the CPTV system, we therefore include recommender systems for both movies and TV series. These recommenders suggest possibly interesting programs based on their content, extracted from WEB sources, and of preferences and previous viewing decisions of the user.
To learn more
I gave an invited talk at the Swedish AI Society meeting in May 2010 describing the CPTV system in more detail. The slides of the presentation are available from SAIS CPTV Presentation.