The Million Playlist Dataset… Remastered

September 28, 2020 Published by Ching-Wei Chen

In 2018, Spotify helped organize the RecSys Challenge 2018, a data science research challenge focused on music recommendation, specifically the task of automatic music playlist continuation. As part of that challenge, we introduced The Million Playlist Dataset: a dataset of 1 million playlists consisting of over 2 million unique tracks by nearly 300,000 artists. This represents the largest public dataset of music playlists in the world. The challenge ran from January to July 2018, and received 1,467 submissions from 410 teams. After the challenge was complete, the dataset and challenge website were taken down, with a summary of the challenge and winning entries published in the ACM Transactions on Intelligent Systems and Technology.

Since then, many researchers have contacted us to request the dataset be made available for ongoing research in music recommendation and machine learning. Today, we are very excited to announce that we have re-released the dataset and challenge on AICrowd.com! 

The dataset and challenge will be available on an ongoing, open-ended basis, and allow for non-commercial, open research use. We hope that this re-release will enable further research and improvements in the field of music recommendation and automatic playlist continuation.