Abstract

If you have ever tried to compose a piece of music from scratch, you will have likely encountered the phenomenon of “writer’s block”. In such a circumstance, it would be useful if you were using notation software that were capable of generating inspiration for the composer. In this talk, I discuss methods for algorithmic melody generation. I begin by building a music-theoretic foundation from physical and mathematical principles, to enable an understanding of pitch and melody in the context of western classical music, followed by an introduction of Markovian techniques as a candidate solution. I then proceed to discuss heuristics to improve the Markov Chain model with respect to musicality, and conclude with a discussion of the larger issues encountered with this approach, including how the Markov property hinders overall cohesion in the composition.