From manual to assisted playlist creation: a survey

Abstract

Nowadays, thanks to the popularization of music streaming services, we gained access to millions of songs to listen to. One of the methods employed by these services to support browsing and promote song discovery are playlists. Additionally, creating and sharing playlists over the Internet have become common practices. A playlist can be defined as a “sequence of songs meant to be listened to as a group”. Research on playlist creation has been done according to three perspectives: i) manual creation; ii) automatic generation and recommendation; and iii) assisted playlist creation. In this paper we review previous research on these three approaches, which we believe are complementary on the subject of playlist creation. We highlight the importance of combining insights from these three perspectives to better understand the current problems and methods, criteria and techniques, and how they complement each other. Furthermore, we identify promising research directions for the three different approaches of playlist creation.