Content recommendation for group members is one of the main uses of the ChatBot: when prompted, the bot asks the internal system for a list of recommended content based on all participant profiles. The recommender system powers gamification scenarios when participants in a room have not yet decided what content to watch.
Classic recommender systems are based in one way or another on Collaborative Filtering, which is OK for large systems but has sparsity problems in small deployments and lacks explainability. The recommenders in Social TV will be based on content and user profiling and matching between those two spaces:
- Content profiling: contents are characterized based on their associated tags as provided by MediaTag, Optiva Media’s tool for automatic tag assignment based on AI analysis of metadata. These are leveraged to create abstract content profiles that can be compared to gauge content similarity.
- User profiling: based on the ØIL (Empty Knowledge, Interest, Like) history of users. A profile for each user is generated considering their most representing metadata from the content (s)he’s watched. Matching users and contents consist of calculating their similarities based on the user profiles and the content metadata.
- Create a unique recommendation for group content based on each user’s recommendation
Other actionable features:
- Quizzes and trivials about movies they have watched to entertain the room.
- Karma points rewards: Users can earn karma points by engaging with the SocialTV system. Karma points can be gained from the number of videos watched, by commenting, rating, sharing, playing quizzes, etc. Karma points ate used to reward users with premium services, discounts… etc.