vendredi 24 Avr 2009, 17h00m

Although I can't blame it, I'm a bit disappointed in how Last.fm's "Similar Artists" radio works. Since it's based on listening statistics, it has nothing to do with actual musical similarity. Rather, it's based on musical "scenes," and thus the results can often be very insular and restricting.

a. For example, if you play similar artists to a Japanese electronica artists like YMCK or Polysics, instead of getting musically/stylistically, all you get is Japanese bands that people listen to because they're weeaboo and have a Japan fetish.

b. If you play similar artists to old bands with a unique style like The Moody Blues or Crosby, Stills & Nash, you don't get similar sounds…you get other nostalgic 60's groups. Once again, they're banded together like this because a group of people only listen to music since it's from the 60's.

c. One of the most egregious examples is with a band like The Polyphonic Spree. You don't get similar artists, you get stuff that inhabits the same "quirky indie scene."

These probably aren't the strongest examples, they're just things that came off the top of my head. I feel like the current Similar Artists system ghettoizes and stifles musical exploration. When it comes to determining musical similarity, I think crowdsourcing might hurt more than it helps.