Automated Generation of User-Tailored and Time-Sensitive Music Playlists

Streaming music platforms have changed the way people listen to music. Today, we can access to millions of songs with a simple internet-connected device. The drawback is that the selection of what to listen is a long, tedious, ant time-consuming process. This is why, nowadays, we choose playlists instead of songs. Unfortunately, since there are thousands of playlists, the selection process can once again be long, tedious, and time-consuming. In this paper, we design a system to facilitate the listening and discovering of new music. The system automatically generates user-tailored and time-sensitive music playlists and proposes a single playlist to play when the user accesses to a music platform. The system understands the user's listening habits by analyzing the low-level features of songs recently played by the user and by using two different clustering algorithms. A novel designed method uses these data to produce a playlist that expands the user's musical knowledge keeping in mind that a good playlist must contain a mix of new and known music and artists. An implementation based on the Spotify API proved the effectiveness of the approach and showed that the proposal might provide benefits to both users (no time wasted to select what to play) and to music platforms (playing of music that otherwise would remain unknown to users).