Sequential music recommendation based on self-attention mechanism and content characteristics

Abstract

Real music recommendation is a challenging problem, aiming at selecting items that users may be interested in from massive data. Recently, the recommendation algorithm based on deep learning has left a deep impression. However, most of the recommendation methods ignore the content features of music itself or simply splice it with other features, which to some extent affects the recommended system performance. This paper mainly introduces a music recommendation method C_Rec based on self-attention mechanism and convolution neural network. It uses a new method to fuse the features of music vocal music, which improves the scalability of the recommendation method. Experiments on the corresponding data sets show that our approach is progressiveness compared with some existing models.