There have been many open-ish projects working with music genres, but nothing on the scale of Musicbrainz or Discogs as far as I know.
- Maintained on an ongoing basis
- Global reach
- Professional-level musicology
This territory is held by proprietary interests, including data vendors like All Music Guide (Rovi) and Gracenote, as well as music service providers like Spotify and Apple Music.
Maybe that shouldn’t surprise me. The great care and patient labor to create such a data set is generally not a good match for crowdsourcing. Crowds can do amazing things, but they are not methodical.
The Baroque entry in DBPedia is an interesting parallel.
DBPedia has a single genre map for the whole world, I believe. (It might be scoped within a language, so that genre relationships don’t span more than one language).
There could be a graph (“genre map”) which is a superset of all graphs, but it wouldn’t be coherent. There would be multiple roots or starting points, the same genre would appear in multiple locations, the same genre name might be used in different ways.
In terms of data modeling, there needs to be one simple change in DBPedia: add an identifier for the cultural context to the the music genre class.