Help with improving the Acoustid database quality

beets is one of a number of application that use the Acoustid database to fingerprinting and identify songs, but there are many other including MusicBrainz Picard and SongKong

I (the author of SongKong) have been looking at the Acoustid database and have generated some reports that show acoustid → musicbrainz Id pairs that are very very likely to be wrong, but these cannot be removed automatically they need to be checked and removed manually, and there are many.

In the spirit of cooperation I would be grateful for any help in removing and checking these, if you have just 30 minutes to spare to fix a few artists that would be fantastic and could make a big dent in this. This is fixing the live Acoustid database so will benefit all applications that use Acoustid

There are a number of reports at http://www.albunack.net/reports.jsp

I think it’s a losing battle:

How does AcoustID even work for collisions to happen? How do we know the minority entry is wrong? What process causes wrong entries to be added?

Is there any clever way to use everyone’s beets library instead? ex.

  1. barcode scan the images with an album
  2. process of elimination (if an album contains good and questionable tracks)

Actually things are alot better now, we have removed alot of duplicates, and also changes were made at Picard to limit submitting of mbrecordings/acoustid if the track length of the song did not match the mbrecording it was linked to.