Computer scientists at the Massachusetts Institute of Technology claim they've developed a computer program that can accurately predict where a song will appear in the charts. The new Echo Nest software was developed by a team including one Brian Whitman, who says: "The computer can go out on the internet and read a bunch of weblogs and record reviews and get a grasp on trends and buzz and hype and that sort of thing. At the same time, it also can listen to the music - do some signal processing to figure out stuff like dominant key and pitch and rhythm and structure."
Dr Whitman, who claims in previous unpublished research to have accurately predicted the Billboard top ten for several weeks, continues: "Hit prediction is possible. What you do is make a global recommendation - instead of recommending something to a person, you recommend it to a very large community and see how well it will react," but adds, "it's not really an interest of ours."
He goes on to stress that this software is not like the collaborative filtering systems used by online stores such as Amazon - that thing where it tells you what you should buy based on your choices and what other people have been buying - saying: "That's a way of saying 10 people have bought this CD and these other people bought this other CD so you should like this other CD, and it works purely on sales. The way we're trying to do things is more organic, from the ground up, looking at communities of people."
Whitman's long term plan is to develop ways of organising people's individual music collections. As their library of digital tracks grows ever larger, says Whitman, listeners will need intelligent automated help in quickly choosing the songs they listen to: "You can talk to [the software]: play me something romantic and quiet, for example. Or you're going jogging and you want something to stay the same rhythm and mix into each other. I want to help people find music, and I want artists and labels to find people".