A Computer Model that Generates Biography-like Narratives

This paper presents an initial decomposition of the process of creative storytelling into subtasks that are relevant for studying where and how creativity plays a role from a computational point of view. Five basic subtasks are identified: building a world to act as setting for the story (including characters, locations, possible actions), generating a set of events that take place in that world, selecting from that set of events those that are worth telling, identifying a particular sequence in which to tell them, and finding appropriate linguistic realizations for each event in that sequence. To test the model, an initial prototype is presented that operates on logs generated artificially by a social simulation built by a multiagent system. A second module addresses the task of generating a textual narrative for a given log. Examples of system input and output are presented, and their relative merits are discussed. The final section discusses future lines of work that may be worth exploring.