M2: Aleatorica: A Mosaic of Popular Interval Content

Valerie Sazonova is an undergraduate student at the University of Alabama pursuing a degree in Computer Engineering and Mathematics. She's interested in exploring the intersection of technical research and creative practice, focusing on bridging computational approaches from music information retrieval and traditional music-theoretical analysis. Her research experience ranges from applying machine learning for radar-based sign language recognition to contributing as a student researcher at NYU's Music and Audio Research Laboratory. Additionally, she has collaborated with artists, designing novel electronic musical interfaces, composing for unconventional instruments, and building self-oscillating sound art installations.

Aleatorica is a mosaic-like music map that visualizes and sonifies the interval content of a large dataset of music by aleatorically regenerating its melody, creating a sonic space for exploring the continuum of melodic features in contemporary music. Following the idea of a 'lecture-concert', this performance will touch on the computational and music theoretical methods used to create such an interface. In order to create the map, a large dataset of MIDI format songs was collected and processed into interval distributions based on note transition frequency. Afterwards, Earth Mover's Distance is computed between distributions using a music-theory informed cost matrix and subsequently used in the UMAP algorithm to generate a latent representation of the dataset. Afterwards, a tiled design is generated using Voronoi tessellation on the dimensionality reduced plot and colored based on quantifiable music theoretical features on the interval distributions. Finally, the interval distribution is used to derive a Markovian transition matrix, which can be used to aleatorically generate music. Additionally, Aleatorica will be available as a website, promoting audience interaction and personal discovery. The resulting performance is a compelling inspection of the landscape of modern melody, blurring the lines between music data visualization and data-driven sound art.