Spotify Music Recommendations Through Deep Learning
Millions of people listen to music every day using the music streaming service, Spotify. Spotify has won over millions of users because of its advanced music recommendation feature. The music recommendation feature on Spotify uses a variety of algorithms in combination with user information in order to make the most effective song suggestions. Currently, Spotify intern, Sander Dieleman has been working on a new way of coming up with song recommendations that incorporates the use of deep learning. Deep learning “is a specialized sort of machine learning, one that involves training systems called artificial neural networks.” If Spotify implements this new system into their current process they will be able to make more effective suggestions and enhance the overall user experience.
This use of deep learning within Spotify would alter their current music network, used to make song suggestions. With the implementation of this new strategy the system would create nodes out of the songs the user has saved and then use these songs to create edges to new recommendations. Through deep learning the songs will be analyzed for traits such as pitch, chord progression, bass, and other musical qualities. With these qualities as the edges of the graph, the system will be able to link to other nodes, that are new songs, with similar traits. Additionally, the system could use positive and negative ties in order to decide whether the suggestion is highly similar or only slightly similar to the starting song. This usage of deep learning is a break through in music recommendation because the system will no longer only recommend songs from similar bands, but give the user songs with similar musical qualities and lyrical style. Once this system is added to Spotify, it will most likely increase the overall user network of Spotify.