Mark Levy (Last.fm) presented at ISMIR his work on tempo estimation using crowd-sourcing. And most of his songs have a clear match in the MSD!
Therefore, here is the matching file, Mark's data can be found here: SPEEDO / PAPER.
The data is simply artist name / title / track ID 1 / track ID 2 / ... tab-delimited. There can be more than one track ID since the matching is based on artist name and title, and duplicates is a known issue in the MSD. Also, we match 4313 out of 4388 songs, the missing ones probabaly have a slight difference in metadata between Last.fm and The Echo Nest and are difficult to track down.
This work is important because it shows how crowd-sourcing can be used for relatively high-level musical task (other than tagging). Also, regarding the MSD, if you're curious about the quality of the tempo field, this paper gives you a possible answer.
Unrelated, "(Don't Fear) The Reaper" is an awesome song. Cheers!
-TBM
- millionsong's blog
- Login to post comments