Evaluating Cognitive Models of Musical Composition
Marcus Thomas Pearce, Geraínt A. Wiggins · 2007
We present a method for the evaluation of creative systems. We deploy a learning-based perceptual model of musical melodic listening in the generation of tonal melodies and evaluate its output quantitatively and objectively, using human judges. Then we show how the system can be enhanced by the application of mathematical methods over data supplied by the judges. The outcome to some extent addresses the criticisms of the experts. We suggest that this is a first step on the road to autonomously learning, introspective, creative systems.