Um modelo de dramatização baseado em agentes cinematográficos autônomos para storytelling interativo

Abstract

Interactive storytelling systems are applications that aim to tell stories and allow the users to interact with the system in order to change the future story events. In respect to the visual representation of the stories, there are several questions related to the manner how the scenes should be shown to the public. Many of these questions that today are faced in the development of interactive storytelling systems have been treated before by several filmmakers. Over the years the cinema has evolved and established various principles and rules to be adopted during the creation of a film. However, unlike movies, where every scene is carefully planned before it was filmed, and if a problem occurs it is simply filmed again, in interactive storytelling this freedom does not exist; usually the storyline of the narrative results from the real-time interaction between the viewers and the virtual actors. The dramatization of stories, whose sequence of events is unknown, depends on a robust dramatization system, which adapts itself appropriately to any type of scene. This dissertation presents a dramatization architecture composed by a set of cinematography-based autonomous agents inspired on the cinematography. To represent the knowledge of the agents is used an approach based on classifiers. More specifically, the architecture uses the Support Vector Machine classifier to classify the best camera shots, visual effects and sound tracks to the scenes.