|
|
|
|
|
|
by kpennell
4338 days ago
|
|
|
Ok idea, but didn't really work for me. I follow quite a few artists on soundcloud. I have several hundred likes. Most of the time, I'm not into the music the artists I follow like. It's simply too all over the place. I clicked next 6 or so times but I didn't like any of the songs given to me. I made a slightly different take on this: http://likesorter.herokuapp.com/# It shows me in a table my likes and then allows me to see (in table form) the likes of artists I follow. Being able to sort my tags, plays, and song likes, I get more signals on whether I might actually like the song. Music discovery is a really tricky problem. It takes roughly a minute each song to know whether I like a song or not. There's almost too many microgenres to keep track of (I like tech house but not trap house). My tastes are eclectic: you'll find cheesy songs right next to the hippest coolest disco funk track. The artists I follow probably have similarly eclectic tastes. |
|
|
interesting "human algorithm" approach, but I believe machines can help sorting the recommendations with a better bird eye view ;)
btw, I am currently working on an algorithmic improvement to discriminate mainstream tracks from alternative tracks if your interested into it