Song Recommender System Based on Emotional Aspects and Social Relations.

Music streaming services have opened the possibility of accessing huge quantities of songs and more sophisticated data can be utilized by recommender systems to improve their performance. Some recommendation methods dealing with different music features have been proposed during the last years, but most of them do not consider emotional aspects. The recommender system presented in this work, allows the classification of the music into emotions from acoustic characteristics extracted directly by means of an automatic analysis of the songs. These emotional aspects of the songs are incorporated to the proposed recommendation models, that also include recommendations to groups of users from their social relationships. The experiments show an improvement in the recommendation reliability obtained by this proposal against the classic collaborative filtering recommendation approaches, for both individual and group recommendations.