Author: Demske, Thomas R.
Title: Recognizing Melodic Motion in Musical Scores: Rules and Contexts
Institution: Yale University
Begun: March 1987
Completed: March 1993
Describes a computer-implemented formal system for recognizing surface melodic connections in selected score excerpts. The system comprises two parts: rules for determining which if either of two potential melodic connections is stronger; and an algorithm which merges such pairwise distinctions into a meaningful whole, whereby some melodic connections emerge as more salient within the musical context. The algorithm is resilient; rules may misfire or clash without skewing the final results. Voice-leading issues are engaged empirically at the note level, and the process of listening is formally cast as a complex, evolving dynamic among competing determinants of melodic activity.
Keywords: voice-leading, melody, AI, perception, foreground, segmentation
2 Assessing Melodic Connections
3 Definitions, Assumptions, Limitations
4 The Line Building Algorithm
5 The Rule Base
6 Applying the Rule Base
7 The System at Work
118 Bishop Street, #E
New Haven, CT 06511