Comment by vulkoingim a day ago
Working on a little project to make Spotify recommendations better.
You get to choose the genres you're interested in, and it creates playlists from the music in your library. They get updated every day - think a better version of the Daily Mixes. You can add some advanced filters as well, if you really want to customise what music you'll get.
Not directly, no. I still rely on the data that Spotify gives me that relates to artists' information; and it's not great a lot of the time. E.g. there are obvious cases where artists belong to genres that they should not belong to. I do have some ideas for improvement, but they are still WIP.
What it allows you to do, though, is create your playlists with extended filters. E.g. you can select genres, and at the same time exclude genres - that helps with the "cross-contamination". You also get a view of all the artists that match your selections and you can add exclusions for them as well. It is a bit of manual work, but it works pretty good for me personally.
Does this deal with, what I call the "Armin Problem?" I typically listen to EDM and there is an extremely popular DJ named Armin van Buuren, and he has other aliases (which exacerbates the issue). His sets end up directly on music platforms, and he pulls in a ton of EDM sub-genres (which makes him a great DJ!). Any recommendation algorithm that visits one of his aliases is doomed to be connected to every other sub-genre, so I might be listening to progressive trance, and be in the mood for that, and end up on deep house (as an extreme example). Within EDM, genres can be as different as blues is to metal.