AudioLens: Audio-Aware Video Recommendation for Mitigating New Item Problem

From the early years, the research on recommender systems has been largely focused on developing advanced recommender algorithms. These sophisticated algorithms are capable of exploiting a wide range of data, associated with video items, and build quality recommendations for users. It is true that the excellency of recommender systems can be very much boosted with the performance of their recommender algorithms. However, the most advanced algorithms may still fail to recommend video items that the system has no form of representative data associated to them (e.g., tags and ratings). This is a situation called New Item problem and it is part of a major challenge called Cold Start. This problem happens when a new item is added to the catalog of the system and no data is available for that item. This can be a serious issue in video-sharing applications where hundreds of hours of videos are uploaded in every minute, and considerable number of these videos may have no or very limited amount of associated data. In this paper, we address this problem by proposing recommendation based on novel features that do not require human-annotation, as they can be extracted completely automatic. This enables these features to be used in the cold start situation where any other source of data could be missing. Our proposed features describe audio aspects of video items (e.g., energy, tempo, and danceability, and speechiness) which can capture a different (still important) picture of user preferences. While recommendation based on such preferences could be important, very limited attention has been paid to this type of approaches. We have collected a large dataset of unique audio features (from Spotify) extracted from more than 9000 movies. We have conducted a set of experiments using this dataset and evaluated our proposed recommendation technique in Yterms of different metrics, i.e., Precision@K, Recall@K, RMSE, and Coverage. The results have shown the superior performance of recommendations based on audio features, used individually or combined, in the cold start evaluation scenario.