Thesystempresented here showsthe feasibility of modcling the knowledgeinvolved in a complexmusical activity by integrating sub-symbolicand symbolicprocesses. Thisresearch focuses onthe questionof whether there is any advantagein integrating a neural network together with a distributed artificial intelligence approach within the music domain. Thesystemis designedto perform in real time and to be used for interactive computermusiccompositionor performance. Thehybrid approachintroduced in this workenables the musicianto encodehis knowledgeand aesthetic taste into different modules.This is doneby applyingthree distinct functions: rules, fuzzyconcepts, andlearning. As a case study, webegan experimenting with first species two-part counterpoint melodies. Wehave developed a hybrid system composedof a connectionist moduleand an agent-based moduleto combinethe subsymbolicand symboliclevels to achieve this task. The technique presented here to represent musical knowledge constitutes a newapproa...
Claudia V. Goldman, Dan Gang, Jeffrey S. Rosensche