Predicting a Viewer’s Mood for Movie Recommenders (2013)
For his graduation project, Tijl Kindt researched whether it is possible to predict someone's mood after watching a movie. For this he used tags that are given to movies in the IMDB and Netflix databases to create an "affective profile" of a movie. When combined with prior-mood information, these affective profiles help predict a viewer's mood after watching the movie. This was shown in an experiment where people reported their moods before and after watching a movie.
Mood-based movie recommending
With an ever-growing easy-access to movies, good recommenders are getting more important. Mood-based and context-aware movie recommenders could contribute by taking a viewer’s mood and context into account, when recommending a movie to watch. Being a film aficionado himself, Tijl decided to work on this problem and developed his own methods for doing so.
Tijl was supervised by both an internal Media Technology supervisor and an affective computing expert from TU Delft. His project was finalized with a good graduation thesis, based on solid research.