Efficient Similarity Based Methods For The Playlist Continuation Task

Abstract

In this paper, the pipeline we used in the RecSys challenge 2018 is reported. We present content-based and collaborative filtering approaches for the definition of the similarity matrices for top-500 recommendation task. In particular, the task consisted in recommending songs to add to partial playlists. Different methods have been proposed depending on the number of available songs in a playlist. We show how an hybrid approach which exploits both content-based and collaborative filtering is effective in this task. Specifically, information derived by the playlist titles helped to tackle the cold-start issue.