I haven’t really talked much about Audioscrobbler, but I’ve an idle moment. The idea is great, an open-source project that monitors the music you play on your computer, and works out which artists and songs you like most. Armed with this data, it can then work out which bands are similar and what songs you might like but haven’t yet heard. I’ve been a member for about a month (no recommendations yet, though they have been busy fixing server holdups recently). But the idea is still neat - in a world where we get so much information thrown at us, having benevolent, helpful spyware doing recommendations for all kinds of media (we could have similar schemes for hypertext and video) is one way of helping refine what we get and allow us to make meaningful choices. They’ve (apparently) reopened the signup process so go join if you’re interested.

Audioscrobbler’s is great, though I can think of some extras I’d like - how about a linkup with The Covers Project so that you can find out remixes and covers of songs you like (maybe that’s just the music geek in me). Also, badly tagged files can muck up the data, but rather than use the slightly heavy-handed method of tagging files by analysing the audio and looking up the resulting fingerprint in a database, why not use the existing title data? You could compare a song’s title in its ID3 tag to existing titles in the database. If it doesn’t match a title exactly, do some some Bayesian analysis comparing it with existing titles and work out which is the correct way to spell it, and re-tag accordingly.