Toward Vignette-Based Story Generation for Drama Management Systems

Drama management is a technique by which a virtual world and computer-controlled characters within the virtual world are manipulated in order to bring about a dramatic experience for the human participant. Generative drama management systems use artificial intelligence to dynamically generate new story content in order to adapt to the participant. Story generation poses many challenges, one of which is ensuring that story content that is generated is good. In the absence of computational models of story aesthetics, we present an approach to story generation that makes use a library of vignettes – short narrative segments – that are presupposed to be “good.” We introduce a story planning algorithm inspired by case-based reasoning that incorporates vignettes into the story being generated. The story planning algorithm requires that vignettes be in the appropriate domain. We further describe an approach for automatically translating vignettes from one domain to another, using analogical reasoning.