m24. Recommendation engine by ExpertSystem

This module recommends videos, using data extracted from comments from users.

During watching a video, in order to improve the number of videos watched per user, several slots with recommended video will be shown on Apple TV GUI. The Recommendation Engine (RE) is based on two main things: User behavior. User is represented by an anonymous ID USER and the following related data : List of video/audio items watched by the ID USER, List of the percentages in relation to the entire video List of feedback (Like/UnLike), Emotions (disgusted, fearful, sad, angry, joyful, surprised) The properties used to predict the best videos for a given user at a particular moment is divided into: Historical profile: from the info collected in the user’s history. Session's profile: from the user behavior in the current session Video characteristics. This could be represented by a set of information already existing in the video database of Deutsche Welle enriched by the semantic tagging provided by semantic analysis of Expert System. The RE needs: A large amount of data on user behavior. This is important to have a story of each user about its behavior. Feedback from users about the response to recommendations provided to allow the RE to learn and improve in an automatic way the quality of recommendation. Result Provided that both user behaviour data as well as feedback on given recommendations can be collected, the RE is expected to give quickly improving results with the help of automatic machine learning mechanisms. Input (ID A/V, ID USER) Output ( list of lists of ID A/V recommended) Contact http://www.expertsystem.com/ for more info