Spaniards develop chart-topping program

A small Spanish company says it can use artificial intelligence technology originally developed for the banking and telecoms industries to predict if a record is going to be a hit or not. It boasts an impressive track record, having successfully predicted hot sales for Norah Jones and US band Maroon 5.

Barcelona-based Polyphonic HMI reckons it has identified 20 elements of song construction - including melody, harmony, tempo, pitch, octave, beat, rhythm, fullness of sound, noise, brilliance and chord progression - which its "HitSongScience" program matches against a database containing 30 years' worth of Billboard hit singles. The database, currently containing more than 3.5m songs, is updated weekly with new releases.

Each song is mapped onto a grid called "music universe" and is positioned according to its mathematical characteristics. Songs with mathematical similarities are positioned very close to one another.

Hit songs have common characteristics, the company claims. It is rare to see a hit song which falls outside its chart-busting "scientific clusters". When this happens, the aberration is often due to the rogue composition's lyrical content. Some hip hop songs have become popular for this reason, as did a patriotic ditty released just shortly after the 9/11 attacks.




Rest of Story: The Register