Music Recommendation Algorithms Based On Knowledge Graph and Multi-task Feature Learning

Abstract

Abstract Both sparsity and Cold-Start Problems are inevitably encountered in the music recommendation scenario. Auxiliary information have been utilized to music recommendation algorithms to offer users more accurate music recommendation results. This paper proposes an end-to-end framework MMSS_MKR, which uses the knowledge graph as a source of auxiliary information to serve the information obtained from it to the recommendation module. The framework exploits Cross & Compression Units to bridging the Knowledge Graph Embedding task and the recommendation task modules. We can obtain more realistic triple information and to exclude false triple information as far as possible since our model obtains the triple information through the music knowledge graph, and the information obtained through the recommendation module is used to determine the truth of the triple information.And thus, the knowledge graph embedding task is used to serve the recommendation task. In the recommendation module, multiple predictions are adopted to predict the accuracy of the recommendation. In the Knowledge Graph Embedding module, multiple calculations are used to calculate the score. Finally, the loss function of the model is improved to help us to obtain more useful information for music recommendation. The MMSS_MKR model achieves significant improvements in music recommendation over many existing recommendation models.