Slicing Up Music: Computational Approaches
Presenter Name:Konrad Swierczek
Co-Authors:Karen Chan, Matthew Woolhouse
Traditional music theory-based analysis often involves a high degree of subjective judgments and variability between researchers: individual analysts differ in how they define individual harmonic or rhythmic events. While this does not present issues for individual analyses, these results may not be reflective of how music is perceived in any population beyond the researcher who performed the analysis. With the increasing availability of complex computational models for music analysis rooted in the disciplines of music cognition and empirical musicology, consistent and objective methods of defining events in symbolically notated music (in this case, western staff notation) are required. More specifically, computational methods with a high degree of generalizability and automation are necessary to facilitate corpus analysis with minimal analyst/researcher intervention. Three such methods are explored: subdivision to the shortest rhythmic event, grouping to a harmonic rhythm or tactus, and reduction of non-structural elements (i.e., figuration). While each of these methods has advantages and disadvantages, integrating all three approaches may yield the most representative output for defining harmonic events in western symbolic notation. The output of these computational methods are presently being applied to perform corpus analysis using psychological models of dissonance, voice leading, tonal attraction, and key finding.