Knowledge Representation and Reasoning of Animated Stories
We have introduced logical formalisms to represent the building blocks of a story world, referred to as domain knowledge. In particular, we introduce an event-centric representation for representing the atoms (smart objects, affordances, and events) of a narrative in a modular, and extensible way. To this end, we have introduced a graphical, hierarchical formalism for authoring narrative atoms, termed as Parameterized Behavior Trees, which can scale to complexity in interaction and number of actors.
Computer-Assisted Narrative Animation Synthesis (CANVAS)
Leveraging the building blocks of stories, described above, we have introduced graphical abstractions that allow anyone (not just expert content creators) to rapidly generate new animated stories, by allowing them to focus on aspects of the story that are crucial to their creative intent, while ignoring tedious details that would otherwise detract them from the story authoring process. This is made possible with two main contributions. First, we have developed a graphical interface for story world specification. Second, we have introduce visual storyboards as a metaphor for storytellers to author their own animated stories. Our platform is accessible to novice users and provides the right level of abstraction for creating story-driven animated content. In order to further alleviate the authoring burden, we introduce automation techniques to help users by automatically detecting and resolving inconsistencies in their stories, and automated story synthesis. Our platform is immediately accessible to novice users, while internally relying on formal abstractions of stories, and the use of sophisticated partial-order planning techniques for story inconsistency detection and resolution.
Computer-Assisted Authoring of Interactive Narratives
We have introduced extensions to the aforementioned logical formalisms to efficiently represent interactive animated stories. In particular, we focus on freeform user interaction where the player has the ability to interact with characters at any point in the narrative, and is not limited to a discrete set of story choices. Our contributions are three-fold: (1) We have introduced Interactive Behavior Trees, which are a novel graphical formalism for authoring interactive stories. (2) We have developed computational techniques for automatically detecting and resolving conflicts and inconsistencies in interactive stories. (3) We have introduced a theoretical framework for analysing and evaluating story complexity.