Séminaire : Assisting Western Popular Music Guitar Practice and Tablature Composition with Machine Learning

Guitarists who play Western Popular Music (WPM, like Rock, Pop, Metal or Funk) are usually little trained in music theory, a great part of their learning being autonomous and informal. To learn new songs or share them to other guitarists, they will commonly resort to audio recordings and tablatures, a form of notation based on gesture rather than music theory. Newly composed songs however tend to emerge through experimentation and jamming, either individually or in groups, with little importance given to musical notation in the process. This seminar will present some computational methods that aim to assist guitarists both in the learning and compositional phases. Modern machine learning approaches can for instance be used to draw insights on the way guitarists learn and compose WPM, by studying large databases of WPM tablatures. Automatic analysis and recommendation models are used to provide detailed feedback on a tablature's difficulty, and recommend learners with appropriate songs, given their current level. We also show how tablature notation software can be augmented by machine learning features to provide suggestions to composers, for instance on how to complete an accompaniment guitar part, where to add expressive playing techniques, or by generating an example of an accompaniment bass track.