“Why Twitter sucks for finding new music”
A continuing debate in which Anthony Volodkin says:
The signal to noise ratio for finding something cool in a content-specific area like music appears to be poor on Twitter right now. Last.fm/Listening to… updates, assorted services republishing data and people’s casual mentions of artist/tracks all mix up the reasons why a given update would appear in someone’s stream.
The tempting approach (taken above and by many others) is to sum up all of this activity and attempt to extract insight. The problem, of course, is that numbers trend to the most obvious uninsightful things and the things of real interest are buried and are easy for a machine to confuse with noise.
Solving that one is a bit more tricky. :)
I get the feeling Anthony’s talking solely from a point of view of aggregating the various mentions across the whole of Twitter into some sort of chart or list of recommendations. But as with any form of recommendation - algorithmic or otherwise - some sort of filtering is necessary.
The Hype Machine’s filtering is built into the medium - people who post MP3s to a blog are likely particularly music-savvy, and so unlikely to only blog about The Beatles or Madonna, as seen in the TweetJ screenshot in Anthony’s original post. Actually sharing an MP3 adds an extra layer of difficulty/effort to simply mentioning an artist in a blog post, filtering out the bulk (the mainstream?) of artists, or “the noise”. If Hype Machine based its stats not just on MP3s posted or linked to, but on overall mentions, I think it would cover a very different range of artists to that which it covers now.
Twitter-based music discovery services like TweetJ, as Anthony correctly points out, “sum up all of this activity and attempt to extract insight”. But that’s where they’re going wrong - they shouldn’t attempt to sum up all of this activity, but only some of it. Should that be the artists mentioned by only a hardcore of A-List music tweeters? Only Spotify links? Only blog links? Tweet Machine takes the route of only tracking links to Hype Machine pages, and I think it goes some way to showing Twitter’s (potential) power as a way of finding new music.
My own personal method of filtering out noise is to follow only people that I know tweet good music. This is human involvement, not machine filtering. But I don’t think the answer is to rely solely on machine filtering. On some level, there needs to be a bit of editorial curatorship, something that stops everyone’s opinion getting turned into a recommendation.
Make it more difficult for a human’s opinion to get picked up by a machine, and only those who truly care - those who typically make good recommendations, one might argue - will have their opinions turned into recommendations.
(Anthony: I’m sure I’m probably teaching my Grandmother to suck eggs here, but it’s all in the spirit of stimulating public debate!)
Please feel free to comment or re-blog with your opinions.
