I’ve written a few times in the past about various attempts to help autonomous systems develop music and other creative works on their own. Most of the time, these systems consume the back catalogue of a particular art form, and then use pattern matching to discern what notes might work together.
Whilst it’s certainly interesting, it’s debatable whether it’s really ‘creative’. New research from the University of the Basque Country takes a similar path in work whereby a machine can automatically generate new tunes based upon a collection of tunes used in a Basque form of song, called bertso.
The project has involved the creation of the BertsoBot to both create new music and classify it autonomously. The team believe that whilst previous approaches have been based upon grammars or statistical models, a more nuanced approach is required.
“The coherence of the melodies will need to be taken into consideration to be able to generate melodies that are easy to understand. We would need to be sure that certain segments are repeated within the new melodies, not only on the note level but also on other more abstract melodic levels,” they explain.
Classification of music
The work has resulted in two new methods for classifying music autonomously. The first revolves around genre and is based upon a new way of representing music by grouping together similar bertso tunes. The project then analyzes the tune and specifies what it’s similar to, or into what genre it can be classified.
These clusters can then be used to generate new melodies in a style that’s similar to others in those clusters.
“We are proposing a way in which bertso tunes can be represented and then how these tunes can be classified. We have come up with a method which can then be applied to another kind of corpus, to another kind of music,” the authors explain.
This has enabled them to automatically create new bertso tunes. Whether this qualifies as being creative in the artistic sense we generally recognize, I’m not so sure.