I lost my will to write here on the last months because I thought there were no great advances, however, I've realized it's not like that.
I've reached an experimental phase: I'm testing my models making people to rank them. The webpage is in spanish, it will be available in english soon. If you want to take a look here is a link: http://lacompositora.elsonidoq.com.ar.
The first experiment is finished and now I'm analyzing the data. One interesting thing, that also looks cool is a way I discovered to graph the probability distribution of a sucession of three notes (two melodic intervals) following Eugene Narmour's Implication-Realization theory.
He establishes some criteria on how we percieve melodic contours, and using that I made a probability distribution.
Within his theory he establishes that there are some criteria on how we organize the melodies. These criteria are cultural-independent, but also he says that breaking some of these rules have some cultural and aesthetic effects.
It's interesting how the probability distribution that my program learns when I serve as input different musical styles are kind of similar, but also different.
Here is one of Tom Jobin (A Felicidade), one of Scott Jopplin (Ellite Syncopations) and one of Beethoven (a dance) in that order. The Y axis indicate the length of the first interval, and the X axis indicates the length of the second interval.