Structural Analysis of Music Using Context-Free Grammar Inference

by
Chester Chan

Abstract:

Determining the structure of a piece of music using traditional methods of analysis requires years of training and practice. Even for experienced musicians, such analyses often require substantial investments of time and effort. In contrast, data compression algorithms can quickly give insight into the degree of hierarchical structure within different kinds of input. "Sequitur," developed by Craig G. Nevill-Manning and Ian H. Witten, is such an algorithm. With one pass through an input, it can generate a context-free grammar by following simple constraints. "Music Sequitur" and "Rhythm Sequitur," which are enhanced versions of the original Sequitur algorithm, are even better at finding structure in pieces because they take commonly-used compositional tools into account in generating grammars. We have found that the degree of structure found by Music Sequitur and Rhythm Sequitur can be affected by the style in which the piece was composed and by the musical form of the piece.

Advised by Richard Ladner

CSE 403
Wednesday
May 30, 2007
3:30 - 4:20 pm