The goal is to be able to have as complete a description of a piece of music as a human being can do. That's beyond the current state of our knowledge. But what we can do now is enough to develop music recommendation systems based partly or totally on the automatic generation of descriptors.

The music world is highly commercial and only the works of the biggest artists are really well known and widely promoted. Something like 10 per cent of music accounts for 90 per cent of music sales, while the remaining 90 per cent accounts for just 10 per cent of sales ? this system could therefore herald a revolution for little-known music and artists.

Though we were originally planning to set up a spin-off company to exploit the project results, we're now looking to collaborate with other firms to introduce our technology into existing products and others that are still in development.

The technique represents a major advancement over the existing methods used by audio software. It improves the way users can organize, navigate and visualize audio files and how they can interact with music on their audio player, PC or the Internet.

The 'since you bought this artist, you might also want to buy this one, as other customers with a similar profile did' method is not that effective, however, because there may be similar songs out there but if they haven't been bought that much they won't appear as a recommendation.

If you didn't have a very accurate description, you're not going to get the identical piece.