Video Recordings

Computational musicology depends on the availability of machine-readable data, and automated transcription is the only viable pathway to truly large-scale datasets. This is the purview of Optical Music Recognition (OMR). In the first part of the talk, I present efforts to build a robust OMR system for recognition of musical manuscripts, in order to make the vast collections of music libraries available for a computational mapping of European music history.  However, there are empty places on this map of music history for which few written sources even exist. One such area is right at its roots: the history of Gregorian chant, especially before exact pitch notation was adopted in the 11th century. In the second part of the talk, I present ongoing research into inferring chant history with computational models. Chant melodies are diverse — practically no two manuscripts record a melody of this tradition in exactly the same way — which is an opportunity especially for phylogenetics. These show remarkable promise in allowing us to plausibly backtrack through the evolution of chant. Thus, taken together, I hope to show how computational tools can read and predict the history of music.