Enhancing the accuracy of ratings predictions of video recommender system by segments of interest

The amount of video content that is available on the web grows at each instant. This fact implicates in an important issue -- video content overload. One way to treat such problem consists on the use of recommender systems. In this sense, this paper proposes a method to enhance the accuracy of the predictions given by video recommender systems by the use of Segments of Interest (SOI). Based on the premise that users tend to like particular segments of a video more than the entire video, and that they are able to mark these segments, these can be used to identify similar people, i.e. the ones who have similar interests about videos. This similarity can be used to enhance the accuracy of the ratings predictions of traditional collaborative video recommender systems. To evaluate this approach, an experimental evaluation was performed. The results showed that the accuracy improvement is directly related with the level of participation of people marking SOI. Thus, as more people collaborate and interact, better will be the recommendation result.