With a track record of 508 songs I finally got my first set of recommendations from AudioScrobbler last night. Topping the list was a group I’d never heard of, Carbon Leaf. A quick search of LimeWire is turning up covers of Whiskey in the Jar, Mary Mac, and Take Me Home Country Road, but their website plays a something a bit more like soft rock. Either way, I’m a little puzzled how this is anything like what I’ve been listening to lately, but the nonlinearity of emergent data is part of what makes it fun.

A few thoughts on AudioScrobbler

It’s a neat project clearly suffering from a lack of resources. I’ve tried a few times over the last year to sign up, and download the appropriate plugins, early this week (just before the Wired article probably caused their load to spike again) I was finally successful, but I worry about the long term health and stability of the servers.

There is a note at the bottom of each that the data is available under a NonCommercial-ShareALike license. Making the data available under a CC license is obviously the right thing to do. After all why would we spend energy building someone else’s database, we all learned better after Gracenote/CDDB, right? What I’m not seeing is any way to download this data in a useful format. I want to see the option for me to download all of my data, and also the entire database (probably anonymized), to experiment with (perhaps inventing a better recommendation engine?)

  • data wouldn’t have to be live, periodic dumps would be fine
  • dump it as text (presumably RDF, maybe as XML), and bzip2 it, and it won’t cause bandwidth issues

I’m using a variant of these instructions for fetching AudioScrobbler recently played info with Magpie, on my about page