Automatic species counterpoint composition by means of the dominance relation

Abstract

This paper introduces a new method for automated composition of first species counterpoint. The method employs the dominance relation – a fundamental notion in the area of multi-criteria analysis, never before used to analyse counterpoints in the context of algorithmic composition research. The dominance relation allows for analysis of a number of evaluation criteria without making any assumptions on the importance of each criterion; this way aggregations of criteria that would lead to loss of information are avoided. Seven criteria are used in this work to evaluate counterpoints in large-scale computational experiments, and the distributions of criteria values are demonstrated for a few types of cantus firmi including popular tunes, Gregorian chants, ascending or descending musical lines, and randomly generated melodies. Mutual discordance of these criteria is also evaluated, revealing pairs of criteria that correlate and others that are conflicting.