On composer attribution

Peter van Kranenburg, Meertens Institute


May 7, 2015. 15.00-17.00 hrs. At the eHumanities group

On Composer Attribution
Recent developments in artificial intelligence enable a new, data-rich,
approach to style-based authorship attribution of musical compositions.
While ‘traditional’ approaches mainly use proof by example, in a
data-rich approach all available musical data can be used, resulting in
statistically more sound attributions and better testable models of
personal style. The major question to be answered when taking a
quantitative approach is which features of the music to employ. These
depend both on the particularities of the personal styles at hand, and
on demands of the computational approach as such. In this talk I will
address this methodological issue, provide an overview of existing
approaches, and present a few cases (organ fugue’s that are attributed
to J.S. Bach, string quartets of F.J. Haydn and W.A. Mozart, early
seventeenth century keyboard music from the Southern Netherlands).
Peter van Kranenburg obtained masters degrees in Electrical Engineering
(2003, Delft University of Technology) and Musicology (2004, Utrecht
University). He developed machine-learning methods for studying
authorship of musical compositions. 2006-2010, he was Ph.D. researcher
at Utrecht University, studying computational modeling of similarity of
folk-song melodies for retrieval purposes. Currently he is researcher at
the Meertens Institute (Amsterdam), involved in the Tunes and Tales project.