In recent years, there has been growing interest in the study of automated playlist generation - music rec- ommender systems that focus on modeling preferences over song sequences rather than on individual songs in isolation. This paper addresses this problem by learn- ing personalized models on the fly of both song and transition preferences, uniquely tailored to each user’s musical tastes. Playlist recommender systems typically include two main components: i) a preference-learning component, and ii) a planning component for select- ing the next song in the playlist sequence. While there has been much work on the former, very little work has been devoted to the latter. This paper bridges this gap by focusing on the planning aspect of playlist gen- eration within the context of DJ-MC, our playlist rec- ommendation application. This paper also introduces a new variant of playlist recommendation, which in- corporates the notion of diversity and novelty directly into the reward model. We empirically demonstrate that the proposed planning approach significantly im- proves performance compared to the DJ-MC baseline in two playlist recommendation settings, increasing the usability of the framework in real world settings.

PDF, HTML

In PROCEEDINGS OF THE TWENTY-NINTH CONFERENCE ON INNOVATIVE APPLICATIONS OF ARTIFICIAL INTELLIGENCE (IAAI-17), San Francisco, USA, February 2017.

@inproceedings{IAAI2017-eladlieb, title={Designing Better Playlists with Monte Carlo Tree Search}, author={Elad Liebman and Piyush Khandelwal and Maytal Saar-Tsechansky and Peter Stone}, booktitle={PROCEEDINGS OF THE TWENTY-NINTH CONFERENCE ON INNOVATIVE APPLICATIONS OF ARTIFICIAL INTELLIGENCE (IAAI-17)}, month={February}, address={San Francisco, USA}, url="http://nn.cs.utexas.edu/?liebman:iaai17", year={2017} }