Composition, Performance and Evaluation - A Dance Education Framework for AI Systems.

Most AI-based dance generation methods consider dance as a selection of movements. While these methods aim for realistic outputs and synchronization with music, domain-specific knowledge related to dance is often not incorporated in these models. This results in outputs that are far from what the dance community considers as high quality dances. Therefore we present CoPE, a dance education framework for AI systems with three key components: Composition, Performance, and Evaluation. By analyzing nowadays dance education, we suggest how to incorporate principles from the dance community into AI-based dance generation systems by aiming for more realistic pieces of art.