Abstract Play: Emotional Sequence Analysis and Improvisation by AI Player
by Shlomo Dubnov
CRCA/Music
University of California, San Diego

We describe a method for play design based on abstract principles of sequence editing, pattern matching and player interaction. The actions of the players consist of improvisations on a template sequence according to notions of anticipation, familiarity and emotion. Sequence learning is achieved using Factor Oracle automation that is extended to allow sequence improvisation and analysis. Since no domain- or mediaspecific knowledge are encoded into the system, the method allows experimenting with generative procedures for creating new content from examples in a variety of time-based or sequentially represented media. Keywords: Emotion, Finite-state machine, Scripting