Audioscrobbler is a plugin for WinAmp and XMMS that collects your listening behavior and aggregates it online along with other users. With data on what people are listening to, the system could begin to connect people to each other, users to new music they’re unaware of, or make predictions on the popularity of a given artist or song.
Unfortunately, the service has yet to reach critical mass (it only has 83 subscribed users), but it appears to be growing slowly. Even with a limited amount of raw data, its experimental artist similarity system pulls out pretty good predictions: a search for slowdive returns Sigur Ros, Mahogany, Bowery Electric and Mogwai, among others.
I’d love to see it grow, and I’d love to get my hands on the data. Mmmm… data.
4 thoughts on “Audioscrobbler”
It says I don’t have enough data yet – but from a casual look, I think you would be one of the members with tastes closest to mine.
ha! i’ve been waiting for someone to be even proximal to me 🙂 i feel like all of the webloggers that have started using it are pretty surprisingly similar, but for some reason i’m out of range. maybe they only update the user similarity map on a daily or weekly basis?
Musicmatch Jukebox does something similar. If you opt in on install (it defaults to opted out) your play stats get aggregated – it uses this to create recommendations. I.e. if you listen to The Dismemberment Plan… they look at their archive playlogs and find out what everyone else who listens to Dismemeberment Plan also listens to and displays those recommendations back to you. — if you use their streaming radio product, it also automatically creates a special radio station based on your aggregated play history.
One difference from audioscrobbler is that MM has millions of participants and play event archives that number in the billions.
You can get a taste of the tech by getting recomendations based on single artists at the link below, but it works even better if you use the product: I.e. if you listen to diverse genres, it can generate a unique list of recommendations just for you.
I’m building a similar data base taken from crawling the p2p networks to generate profiles and then using a data mining technique to generate relationships between artists. Data from a run a while ago is available at http://www.musicbrainz.org