Glyph is worse than some and better than others. He believes that life is just one damned thing after another, that only pop music can save us now, and that mercy is the mark of a great man (but he's just all right). Nothing he writes here should be taken as an indication that he knows anything about anything.

Related Post Roulette

13 Responses

  1. Avatar Richard Hershberger says:

    The classical music (broadly defined) region strikes me mystifying. It doesn’t help that it seems a bit unclear about the difference between a composer and a performer. It reminds me a bit of trying to use Pandora for classical music, where the algorithm may or may not work for popular music, but for classical the results are simply incoherent.Report

    • Avatar Glyph says:

      The makers also appear to be having us on once in a while. For example, Guided by Voices appears under “lo-fi” (so far so good), BUT also appear under “gbvfi” – it’s possible that GbV are influential and prolific enough, that they warrant their own genre, but…

      There’s also one called “Deep Indie Rock”, and I have never heard of ANY of them; I mean not one.

      Admittedly, I’m old and stuff, but come ON…some of these HAVE to be made up.

      Still, it’s a fun toy, and I’d wager that picking something you like in the pop/rock/dance/etc. spaces, and then checking out the proximate artists there, would probably net you a few finds.Report

      • Avatar Chris says:

        Oh, I think I’ve actually posted some Boy Eats Drum Machine here!

        I’m young. I’m young!

        (I couldn’t find anyone else on the “Deep Indie Rock” list that I’d heard of.)


      • Avatar Richard Hershberger says:

        Oh, it clearly works much better for pop/rock. And that is fine. I am mostly commenting on how people come up with these music algorithms and they conclude that because it works well for both rockabilly and grunge, it clearly is universally awesome. In particular, classical music has enough residual cultural prestige that they want to imagine their awesome algorithm works for classical, too. But they aren’t actually interested in classical, and there is no money in it, so no one puts in the effort to come up with an algorithm that actually works for classical.

        That is fine. You know what works for me? Radio stations. I am fortunate to have a very good local classical station, and there are many others that stream online. I will pull up a station from, say, Hong Kong to hear what they play. Life is good.

        But stuff like this is almost comical. I punched in Perotin. He is an interesting test case because he wrote around 1200, and is among the earliest composers where we can actually play it and ascribe it to this guy. It is so early that a lot of Western music ideas of harmony and melody had not yet gelled, and so he doesn’t really sound like what we think of as Western music. So what did this site do with him? It’s reaction was “fuck it: toss him in with everyone else from before 1600 or so” as if nothing had changed over the previous four hundred years.

        As another test I put in John Philip Sousa. It classified him as classical, romantic, and marching band. Going into romantic, he is near Vivaldi, which is weird. Sousa you can put in romantic based on his dates, but under no scheme is Vivaldi plausibly romantic. Going into marching bands, he is in the middle of a bunch of marching bands: Marine Corps Band, various college bands, and so forth: sensible so far as it goes, but not actually useful if I was wondering about other composers to marches.

        Then I put in Stephen Foster. The site has never heard of him. That is just pathetic. That’s when I quite testing it.Report

        • Avatar Chris says:

          (Now Glyph’s young too!)

          I wonder if, with classical music, it’s the algorithm or the input. I don’t know exactly how their algorithm works, but I’m assuming that with their Genome project attributes, they have a big multi-dimensional space and similarity scores based on proximity. If that’s the case, it’s likely the attributes, that is the dimensions, that are the problem, either because the 400 or whatever attributes they use don’t apply as well to classical music, or because the people who applied them don’t understand classical music well enough to assign them to pieces accurately.

          I’ve always noticed that with both jazz and classical, Pandora stations are basically just “any classical” or “any jazz” (though it does seem to make some really broad jazz distinctions).Report

        • Avatar Glyph says:

          If I had found this using it (instead of on Dangerous Minds just now), it would automatically become the GREATEST orchestral app ever: