Though I signed up over a year ago, it’s only in the last month of so that I’ve begun to really use the fantastic Audioscrobbler. Basically, a discrete plugin which keeps track of what music you’re playing (in iTunes, WinAmp, WMA9, XMMS &c) and then aggregates that on a personal page at audioscrobbler.com (here’s mine). I think it uses Musicbrainz – the ‘community music metadatabase’ – as a ‘back-end’, such that clicking on a song will fairly accurately show who else is a fan of that song. As Phil Gyford notes, this comparison ability is darn good for discovery. The author, Richard Jones, is now adding some social software-like features: friends, groups etc. It’s very cool.
Of course, you could use this notion of ‘friends’ as part of a music recommendations system, as I mentioned recently, in addition to the aggregation of broad patterns via collaborative filtering. A potential limitation with Audioscrobbler’s implementation of friends at the moment is that it wouldn’t enable me distinguish a friend whose musical taste I trust (and would want my recommendations to be more weighted by) from a friend whose musical taste I either don’t trust or just don’t want to affect my recommendation. With some of these minor modifications, a more complex recommendation model is within reach though, built around combining collaborative filtering with a way of weighting the scores from ‘friends with similar or interesting tastes’. With music (and perhaps other similar cultural fields), in any given social group (whether <5, <50, or <150), there’s always those whose job it is to make the mixtapes, to DJ at parties, to share their playlists, who others turn to for music recommendations. Combine that complex local knowledge with the blunter tool of ‘the math’ (as Lucas Gonze notes in the comments at Many2Many) by passing collaborative filtering over these enormous aggregations of data from systems like these, and we could be on to something.
Leaving aside the patterns in other people’s ‘scrobbling for a moment, as my Audioscrobbler page developed over time I started observing my listening. And now my obsession has got to the stage where I’ve been ‘teaching’ it, ‘seeding’ the system with a fair representation of what I’m into by leaving a ‘Top Rated’ playlist running overnight in iTunes’ party shuffle mode. This is quite odd behaviour if you think about it – playing music not to listen to, but to ensure that Audioscrobbler has a decent understanding of who I am, musically. As Tom Coates astutely pointed out, it’s another aspect of presentation of self. Which means it’s quite important (as well as being not a little self-regarding, obviously 😉
It’s got to the stage where I’m actually slightly disappointed to think that all the listening I do out on the street or cafe with my iPod isn’t being recorded by Audioscrobbler. So it’s only a part-presentation of self – only part of me is being recorded/transmitted here – my listening at home.
Actually, here’s the opportunity for Apple.
Obviously, across their connected music infrastructure of iPod, iTunes, and iTMS they’re tracking all those play counts too. iTunes could do everything ‘Scrobbler’s currently doing – the networks, the artist lists, the groups, the friends. You could have your personal page at Apple (or rather, within iTunes), keeping a track of your most-listened to artists/tracks – and then publish it within that infrastructure. Then simply extend that with some relatively straightforward features to enable adding friends and localised groups. Again, add some indication of how much you want particular friend’s listening habits to affect your recommendations (i.e. ‘let David affect my recommendations this much; Matt this much etc.) Then mix up with the massive collaborative filtering potential based around a distributed world of listeners (‘spyware’ issues perhaps, ones that Audioscrobbler neatly sidesteps). Moreover, Apple have a post-browser application which could handle presence too (via integration with iChat) – akin to what iChatStatus already does, but within iTunes – people listening to the same track right now; your friends and what they’re listening to etc.
(Obviously, this will work better – or actually work – once Apple have ‘fixed’ their feature of play counts and ratings from your iPod not actually being synched if you manually manage your tracks and playlists. Sort it, Apple!)
[ Another thought on the new iTunes, while we’re on it: shared playlists ahem iMixes – absolutely fantastic feature, but why not do an Amazon Associate-style cut for those whose playlists get bought? That’d be a great incentive. But the publishing of it is fantastic right there, with huge potential. Another more subtle feature is that of enabling playlists to be based around playlists – it’s almost getting SQL-like power here i.e. do a select from ‘ECM’s :rarum series‘ playlist where rating is five stars and date added is in last 3 months &c. ]
So, there are huge opportunities for Apple ahead, if they want to go that way. Though what ‘Scrobbler does of course is accept, aggregate, and publish all that detail cross-platform, from multiple plug-ins/media players, and then make that super-accessible over http/html – which is very smart (and yet, it does hamper it when it comes to denoting presence, and browser-based solutions just can’t match the kind of quick music browsing/searching that iTunes/iTMS was built for, due to its post-browser incarnation).
But this post started about features available right now, and within Audioscrobbler. In fact, it’s currently something of a victim of its own success, creaking under the strain a bit, so I’m somewhat loathe to point more people at it. However, part of the reason I’m writing is to suggest people donate to Audioscrobbler!. They need another fast database server, and I reckon that’s a good cause. Paypal etc will do it. Whatever Apple do with iTunes, Audioscrobbler is out there breaking new ground right here, right now.
Leave a Reply