Using and extending the Tonal Diffusion Model to study Gabriel Fauré's harmonic language.

My aim is to try and replicate this work by evaluating the Tonal Diffusion Model (Lieck et al., 2020) on a new corpus of 98 pieces written by Gabriel Fauré, whose music is poorly represented in the Tonal Pitch-Class Counts Corpus. Moreover, the well-known subtleties of his harmonic language, the modal colors (still, within a tonal framework) that characterize its later manner, may not fully be captured, I humbly thought, by the current model, and may reveal the need for extensions when studying this sort of music. Thus, my second objective is an attempt at exploring some of the ideas suggested by the authors of the TDM, to extend to the model. Drawing the code they provided, I added a dimension to the Tonnetz, by including the interval of tritone; I also implemented a new model class allowing for multiple tonal origins. My intuition is that even minimal extensions such as these could allow to better model and understand the music of Fauré and its evolution. Indeed, Reading R.C. Tait’s dissertation on Fauré (1984) inspired me to develop several hypotheses that I would like to verify by analysing the parameters learned by the TDM (weights associated with transitions on the Tonnetz, path lengths, tonal origins). In particular, I have chosen to adopt his division of Fauré's work into three major periods (ranging respectively from 1861 to 1884, 1884 to 1906, 1906 to 1924).