
Algorithmic Composition (III)
Michael Scott Asato Cuthbert teaches this class session from MIT's 21M.383 Computational Music Theory and Analysis, part of a unit on algorithmic composition. The session opens with questions on the current problem set, then moves into major scales and the use of roman numeral analysis in tonal harmony, including places where roman numeral labeling produces inconsistent or ambiguous results. Cuthbert demonstrates a computational, AI-based system for generating roman numeral analyses automatically, testing it against Mozart piano sonatas to show where the algorithm matches traditional human analysis and where it diverges. The lecture sits at the intersection of music theory and computer science, treating harmonic analysis as a problem that can be formalized and automated rather than only performed by ear and eye. Useful for students interested in how symbolic music representation and pattern recognition can be applied to classical repertoire.