Literature Research into Natural Language Generation for the Virtual Storyteller

This paper presents a literature research into generating stories in natural language based on events created by the Virtual Storyteller, a multi-agent system that simulates actors to create a story. The goal was to come up with recommendations for the system design of an NLG system that can create these stories. This system will be based on an existing subsystem of the Virtual Storyteller called the Narrator. The Virtual Storyteller communicates with the Narrator by means of a story data model called the Fabula model, which contains the events of the story. With the people currently working on the Virtual Storyteller we formulated a recommendation for a new format for the Fabula model based on GraphML in combination with an OWLontology. I will try to make the generated stories more lively by adding focalization, i.e. telling the story from one character’s perspective, and temporal styles, such as flashbacks. Focalization will be done by performing data selection in such a way that only the events the focalized character can perceive will be present. Temporal styles will be made by ordering the content in a non-chronological way that matches the specific style.