A new set of genre tags for MusicBrainz recordings

[alastairp]

Hi everyone,
Late last year I spent some time adding some more genres and tags to recordings in MusicBrainz.

A quick overview of what we did:

• Added 6 Million individual tags to over 1.3 million recordings
• Data aggregated from three different sources
• Data automatically mapped to existing MusicBrainz tags, and manually verified
• Source code and data for this project at GitHub - metabrainz/genre-matching: Experiments to match external genre datasets to musicbrainz genres
• Tags added by three bot accounts, Editor “tags-lastfm” - MusicBrainz, Editor “tags-discogs” - MusicBrainz and Editor “tags-tagtraum” - MusicBrainz

There is some more details about what we did in the source repository, but I’ll repeat some of it here.

We collected this data from three sources - genre and style annotations from discogs, tags from last.fm, and tags from beatunes. This data comes from an earlier project which I did a few years ago to collect some more detailed genre metadata for submissions that we had in AcousticBrainz.

The tag names in these data sources didn’t always match with what was in MusicBrainz, so we performed some automatic matching to try and match similarly written tags. @reosarevok did a great job of then manually verifying these matches, and in some cases creating some new genres in the MusicBrainz database.

These tags act just like any other tag in the MusicBrainz database, that is, they can be upvoted or downvoted as usual. If you want to upvote something then please feel free to do so. Likewise, if you see something that you don’t agree with, downvote away. You can see a list of the tags applied at

• Editor “tags-tagtraum” - Tags - MusicBrainz
• Editor “tags-lastfm” - Tags - MusicBrainz
• Editor “tags-discogs” - Tags - MusicBrainz

In the case that you find some really suspect tagging choices then let us know and we’ll see if we can make some corrections.

I hope you find these tags useful - the ListenBrainz development team already has some ideas about how to use them to start building some new recommendations to listen to. Let us know how you plan to use them!

[PierPiero]

Congratulations @alastairp

@alastairp

Let us know how you plan to use them!

I would like to group recordings’ tags at the artist level using prevalence criteria, in order to get recommended artists like in Discogs.

Furthermore, a hierarchy of subgenres like in Sonemic would be very useful to drill down into a more specific selection following a preliminary recommendation.

[aerozol]

@alastairp

Let us know how you plan to use them!

No plans yet, but I can’t wait for Listenbrainz to use them in reports etc

Less interested in recommendations but all of this knock on stuff from including genres is great.

Genre tags is one of the few ways to give masses of music data a more human-readable ‘meaning’. In fact the only other one I can really think of is ‘country’
If I stretch for it maybe you could do something with avg. BPM or something as well… ANYWAY, this is so important and thanks so much!

[PierPiero]

@aerozol

In fact the only other one I can really think of is ‘country’

I agree 100%: having a hierarchy by area and genre could be useful in the search results, pending development in ListenBrainz.

[sammyrayy]

this is super cool, thank you for creating these!! the only problem with using last.fm, though, is some artists are purposely given the wrong tag, typically for genres like metal and hardcore. like for example, one of the top-tagged artists for brutal death metal is paris hilton, and the top track is “never gonna give you up”. it’s pretty funny, tbh, but maybe a filter or something could be implemented for the hardcore genres?

[UltimateRiff]

that is pretty funny~

for what it’s worth, we’ve already got that issue, especially with generalized genres like techno, dubstep, and the like. since it’s just one or two votes, just downvote those tags.

in fact, this might be an easy fix, since you can see what each import added by looking at the user pages above. I’ve already removed them from Never Gonna Give You Up. (that’s not to say it’s a quick fix, but it’s easy)

also, just wanted to share Peppa Pig, also tagged as Brutal Death Metal.

[aerozol]

@UltimateRiff

just wanted to share Peppa Pig , also tagged as Brutal Death Metal.

Not anymore, I couldn’t leave a WRONG tag in there sorry

I also removed ‘grown up-music’. Made me chuckle!

[alastairp]

Thanks for the reports on weird tags, everyone. As @UltimateRiff and @aerozol point out, because these tags were only added by maximum 3 accounts, a downvote by enough people is often enough to “fix” the problem. However on this specific case of our friend Mr Astley I will go through and remove this tag from the lastfm user. This is of course a limitation of allowing anyone to add any data to the database, but hey, it’s worked well for MusicBrainz so far, and the wisdom of the crowds certainly seems to be winning this race!

[UltimateRiff]

just out of curiosity, are these bots still adding genres for newly added releases? also, somehow I thought Bandcamp was included here, but I guess not…

[reosarevok]

Bandcamp was being added by a separate script ran by a user (I’m not sure if @MrClon is still running it). @alastairp could probably say whether these other projects are still being added or it was a one-off (I suspect the latter).

[UltimateRiff]

that explains it, it was an AcousticBrainz project. that’s why I couldn’t find the thread~ thank you~

it’s this thread, for those looking