Procedural Narrative Generation in Video Games
Ben Kybartas , Master's student

Area: computer games

Abstract:

Using procedural narrative generation in video games provides a flexible way to extend gameplay and provide more depth to the game world at low cost to the developers. Current examples of narrative generation in commercial games, however, tend to be simplistic, resulting in repetitive and uninteresting stories. In this talk, we will present the challenges and benefits of narrative generation. Our system, called ReGEN, is developed for narrative generation using a context-aware graph rewriting framework. ReGEN uses a graph representation of the game world to create narratives which reflect and modify the current world state. Using a novel set of metrics to evaluate narrative quality, the approach is validated by comparing the generated narratives to other procedurally generated stories, as well as to authored narratives from commercially successful and critically praised games. The results show that our generated narratives compare favourably to authored narratives. ReGEN's metrics provide a new approach to narrative analysis, and the system provides a unique and practical approach to story generation.