Authors: D.B. Carvalho, E.G. Clua, E.B. Passos and C.T. Pozzer
In this work we present an emergent story generation system based on wrong, but coherent, character behavior. The goal is to make them perform more realistically by the use of perception simulation. We use a physical description of the characters and objects that compose the virtual environment, where the story takes place, with a pattern matching process that may result in a wrong observation or interpretation of the world. With this system the characters do not need to be omniscient and act based on their partial vision of the world. Our system was tested to generate a story similar to Little Red Riding Hood, which is famous for its perception failure action, and presented interesting variations for the ending.
Keywords: storytelling; character knowledge base; character perception; character behavior