As the domain tagatune.org has gone offline, with kind permission of the original authors, we now host the MagnaTagATune dataset at City University. The data was collected using the TagATune game and music from the Magnatune label. Credit for collecting this handy dataset goes to Edith Law, Olivier Gillet, and the authors below. If you use the dataset please cite the following paper in your publications.
Edith Law, Kris West, Michael Mandel, Mert Bay and J. Stephen Downie (2009). Evaluation of algorithms using games: the case of music annotation. In Proceedings of the 10th International Conference on Music Information Retrieval (ISMIR)
- Clip metadata in CSV format
- Clip metadata as MYSQL database dump
- Audio features extracted by The Echo Nest API 1.0
- Tag annotations collected from game players in CSV format
- Similarity data in CSV format
- Audio data: mp3 32bps file1(1GB) file2(1GB) file3(1GB)
- Python files to generate the dataset by Olivier Gillet
The dataset has been used in a great number of publications, and more data is available through the following sources
- Numerous web APIS such as the Echo Nest, and Last.FM provide further tags to the songs
- The Magnatune Catalogue labels most of the used songs with 1-3 genre annotations
- Our ISMIR2012 publication allows for using the similarity data with Matlab
- Paul Lamere’s blog introducing the dataset
If you have any further datasets related to MagnaTagATune or comments please email us.