Commentary on “Musical harmonization with constraints: a survey”
摘要
Our 2001 survey on automatic harmonization appeared at a moment when tonal music offered an unusually clear test bed for constraint programming: explicit rule systems, finite domains, and musically meaningful combinatorics. Its citation activity turned out to be longer than we expected, even though we were neither the first nor the last to study the problem. Its most durable insight was the distinction between rule satisfaction and musical interest. Subsequent research confirmed that generating harmonizations that are legal is only part of the problem; representation, ranking, interaction, and style are at least as important. We also suggest that the role of constraints has evolved, from being the whole engine of a harmonizer to becoming a language of control inside hybrid symbolic, statistical, and neural systems.