The approach to genres taken by MusicBrainz seems to be somewhat problematic:
- The curated list of genres seems to have some generally accepted genres missing and some genres listed that are generally accepted as not really being genres (like Steampunk). There is no edit history to show how the spurious genres were raised (though I assume that requests for changes to the genre list are through STYLE tickets here so potentially there is some sort of audit trail).
- There are no descriptions associated with genres - so it is difficult to understand what the genre is, and to understand the differences between closely related genres.
- There is no hierarchy to show sub-genres (single parent) or fusion-genres (multiple parents). So an object can be entered as Romantic Classical with or without Classical. Without a hierarchy, parent genres cannot be implied, and searching by genre then becomes much more difficult.
- Genres are entered as tags, and providing that they are entered exactly as listed in the genre (case independent) then they are recognised as a genre, otherwise they are treated as a freeform tag. This A) does not identify synonyms (i.e. use of Steam Punk instead of Steampunk), B) is highly prone to typos, and C) results in a variety of different typographical cases in the genres displayed / provided to picard.
- Genres are optional - whilst I accept that the lowest level genres may be subjective or difficult to determine and so should not be mandatory, even the highest level genres (e.g. Classical / Jazz) are optional. This leads to highly variable data quality.
- Whilst genres added to releases, RGs, works, recordings, artists etc. are subject to plus1/minus1 ratings, that is a very different approach than subjecting them to the normal MB voting process. Again this leads to highly variable data quality.
There are undoubtedly other tickets suggesting improvements to MB's genre handling (and I will attempt to link them once I have created this ticket), however I have created a new ticket to register a suggestion that MB partner with, and use the Genres data from RateYourMusic.com.
RateYourMusic has an existing, and apparently comprehensive hierarchical genre system - with associated descriptions, curation processes and discussion forums - that is more comprehensive than MBs. Reusing this data as a starting point would provide a fast way of creating an entirely new genre hierarchy with descriptions. Linking to their genre data to stay in step as they process changes, would allow MB to benefit from their curation process (the downside being that MB would be reliant on their process rather than being fully in charge of its own data).
RYM also has some works / recordings tagged with genres that is likely to be reasonably accurate (because their genre linking is- I think - curated) - and providing that we can associate their works / recordings with equivalent ones in MB, it might be possible to import / link this data for a step improvement in genre quality.