Date of Award

Spring 1-1-2025

Document Type

Dissertation

Degree Name

Doctor of Philosophy (PhD)

Department

Music

First Advisor

Quinn, Ian

Abstract

This dissertation trains deep learning models for the Roman numeral analysis and applies these to the analysis of a corpus of over 12,000 movements of Classical music composed between 1700 and 1900. The Roman numeral analysis models obtain performance that is, as of this writing, state of the art. At the same time, this is the first work of scholarship to employ such models at scale for a music-theoretic corpus study. Consistent with music-historical expectations, this corpus study shows that, along many dimensions, such as the distribution of keys employed or the proportion of chromatic scale degrees, harmonic complexity rose steadily over the study period, while along some other dimensions, such as the transition tendencies of Roman numeral degrees like I and V, harmonic complexity fell before rising again. We also observe that major-key practice changed more than minor-key practice, generally becoming more similar to minor-key practice, and that the usage of keys, scale degrees, and harmonies, tended to become more dispersed over the course of the nineteenth century. Finally, the corpus study reveals some striking symmetries between the modulatory behaviors of major and minor keys.

Download

Share

COinS