See code changes in pull request #3152 submitted by anshg1214.

I believe the idea for the genre explorer is more for the visualization of genres, hopefully using both the structured data and similarity data (similar to the music neighborhood feature linked in the ticket) to see the connections. tho clustering genres could be interesting too, especially more for the general user~

UltimateRiff yes, the original framework is here:

https://github.com/regorxxx/Music-Graph

And it can be applied to any database as the user showed in the fork.

I use it to calculate genre similarity and it relies in the idea that genres should be part of bigger categories and interact via, "style clusters" "influences" and "counter-influences".

Every type of link modifies the distance, and the distance between 2 given genres (in fact any type of node) can be calculated with path-finding + some modifiers found along the path accounting for influences.

The HTML representation just uses a simplification accounting for the distance between the direct link of every pair of nodes.

You have a smaller scale but full featured database here:

https://regorxxx.github.io/Music-Graph/Draw%20Graph.html

The ListenBrainz counterpart is missing style clusters and other types of links. It's just a full dump of the database and using every key as genre node. They should be classified into superGenres - group of genres- (rock & roll superGenre), clusters - group of superGenres - (rock genres cluster), musical categories (pop & rock), etc. Finally single genres should be related to other nodes via style clusters which include genres from different superGenres groups (like Neo Folk including Pagan Folk, Industrial Folk, etc.)

But that's up to you how you would want to implement it.

The real aim of the graph is the genre similarity calculation, not the display part though. But sure you can use it for the second thing too.

also adding, there is a similar feature build around just MusicBrainz genre data in another project, called Music Graph. they've also got graphs for other databases, currently Discogs, Rate Your Music, and AllMusic.

https://username116.github.io/Music-Graph/Draw%20Graph%20MusicBrainz.html

I have done some basic mockups for this - I definitely think the current iteration of Music Neighbourhood is an excellent springboard for more stuff like this!

I've kept these mockups (hopefully) simple to implement. There are lots of alternate, more advanced, versions that could be mocked up, like:

• either displaying both (with a different connecting line colour?) or allowing for toggling between genre relationships and user-data genres (as UltimateRiff has suggested)
• displaying both artists and genres in the same 'web' display so you're seamlessly clicking between them
• and/or allowing the user to shift between genre and artist displays at any time (e.g. thinking of them as 'layers' of the display, and the user can move between layers. It would be interesting to think of this like a literal map, where artists almost have a 'position' on the genre map)

But it's dependent on what fits nicely into the code (rather than being forced in), and what whoever picks this up to implement wants to do. I am always happy to do more mockups on request!

Basic genre mockup:

New elements:

• genre search bar at the top
• genre sidebar has blurb from relevant MB genre wiki link + [genre] radio link/image collage (if image collage is hard to do I can do a 'radio' icon instead) + top artists/albums/songs (pick one, or togglable)
• Either using user-data (e.g. users who listen to x genre also listen to x genres) or MusicBrainz data relationships. Using user data is more ListenBrainz' style, but I promise it is going to give weird results haha (electronic and rock everywhere once again...), or both.
• For colours, maybe the gradient of the top two albums

We can link between artists and genres by:

1. Having a extra section in the sidebar that connects to genres, from artists:

2. Display relevant genres in the artist web itself (and relevant artists in the genre web):

Again, I think there are cooler ways to do all this, particularly in terms of 'interleaving' all the data, so please don't feel limited to this, but hopefully these mockups are a good baseline for implementation or further discussion.

also, I don't know whether this part would work, but if we can get genre data from SoundCloud and other similar sources, perhaps we could supplement with that? I don't know if we've got easy access to that, or if we can even use the data, being an open data project as we are... figured I'd throw it out there at least~