Music Similarity: Raw Data and Statistics

As part of our research into music similarity and music recommendation systems, we have been looking at the problem of ground-truth information and how to collect subjective opinions about music similarity. We are committed to sharing this information with other interested researchers, and these pages provide access to the data we have collected and processed. Where practical, we provide data in its most neutral and most useful form i.e. after regularization and tokenization, but before distillation to specific forms we use.

Currently, we have descriptions for the following data sources:

• Musicseer survey data - Results of the web-based survey/game we conducted in 2002.
• Expert + Erdös data - Artist similarity mined from the "Similar Artists" lists of All Music Guide, and extended via hop-counts.
• Playlist data - Playlists retrieved from The Art of the Mix website, parsed to find co-occurrences of particular artists.
• OpenNap user collections - Data on personal collection co-occurrence, measured on the OpenNap peer-to-peer network in 2001 and 2002.

There are also some ancilliary resources:

• Descriptions of the various metrics we defined for comparing similarity measures against ground-truth, and against one another.
• aset400 is our list of 400 current US pop music artists about whom all the ground truth has been collected.
• uspop2002 is the definition of a corpus of 8764 tracks from the aset400 artists that we have used in our recent acoustic-based similarity experiments. We will distribute features derived from these 8764 tracks given enough interest.
• Some notes on text normalization of artist, album, and track names.

Papers

Here are some of our publications that relate to this data:

M. Mandel, D. Ellis (2005). Song-Level Features and Support Vector Machines for Music Classification

Proc. Int. Conf. on Music Info. Retrieval ISMIR-05, London, September 2005. (6pp)

A. Berenzweig, B. Logan, D. Ellis, B. Whitman (2004). A large-scale evaluation of acoustic and subjective music-similarity measures

Computer Music Journal, 28(2), pp. 63-76, June 2004. (14pp)

B. Logan, D. Ellis, A. Berenzweig (2003). Towards Evaluation Techniques for Music Similarity

White paper (keynote!) at the Workshop on the Evaluation of Music Information Retrieval (MIR) Systems at SIGIR-03, Toronto, August 2003. (5pp)

A. Berenzweig, B. Logan, D. Ellis, B. Whitman (2003). A large-scale investigation of acoustic and subjective music similarity measures

Submitted to ISMIR-03, Baltimore, October 2003. (8pp)

A. Berenzweig, D.P.W. Ellis & S. Lawrence (2003). Anchor Space for Classification and Similarity Measurement of Music

Proc. ICME-03, Baltimore, July 2003. (4pp)

D. Ellis, B. Whitman, A. Berenzweig, S. Lawrence (2002). The Quest for Ground Truth in Musical Artist Similarity

Proc. ISMIR-02, Paris, October 2002. (8pp)

Acknowledgment

This material is based in part upon work supported by the National Science Foundation under Grant No. IIS-0238301. Any opinions, findings and conclusions or recomendations expressed in this material are those of the author(s) and do not necessarily reflect the views of the National Science Foundation (NSF).

Last updated: $Date: 2005/11/16 19:16:52 $
Dan Ellis <dpwe@ee.columbia.edu>