Digital Media

MATMUSIC: Mathematical modelling of  musical hits

What mathematics underlie a musical hit? Is there any secret scientific formula to predict it or does it depend on random existing conditions? Is there an absolute definition of musical hit? The main goal of this project is to devise a software tool to predict the chances of success for a given song. The approach adopted is to apply information-theoretic measures to song features such as spectrogram, melody, tone and tempo. The tool may be used as a decision support system for the recording industry.

In music, we often think that the number of possible melodies can be infinite. Strictly speaking, it is true. However, the set is drastically reduced if we focus on melodies 'acceptable' for humans. Furthermore, if we focus on popular music we find a surprising fact: we can generate most of the songs created in 30 years by a few combinations of mathematical patterns. These patterns are not fixed but are clear enough as to follow relatively simple probability distributions.

Entropy-magnitude as the information contained in a data-flow is key to the process of creating a musical work. Researchers say it is impossible to predict if a song is 100% successful because not only depends on the work itself but on how it promotes and many other environmental factors. But there is evidence that the mathematical ingredients to create a really successful song are there. "Somehow our brain 'knows' and our task is to translate it into mathematical language”, says the principal researcher. The song eventually will be a 'hit' or not depending on the interaction of the work with the environment.

The environment is not just the public. There are more factors: the artist or group, the label, the consistency of the lyrics with the artist, the production and arrangements, etc. So the key is to separate the phases by which a song must pass to reach the public. The first phase is the mathematical structure of the theme. The second phase is the part of the interaction with the environment.

One of the objectives of the project is to simulate a composer. Suppose you want to compose a song in the style of Coldplay and we would like to predict what kind of song would be capable of being made by this British band. Well, in principle, this would be possible. Just figure out what combinations of basic patterns are preferred when composing their songs. According to the researchers, this opens a fascinating research field. If the study is successful, you would be able to compose a melody that merge, for example, the type of composition of the U2 AMARAL and hear how it sounds.