The Need for Multi-Aspectual Representation of Narratives in Modelling their Creative Process

Abstract

Existing approaches to narrative construction tend to apply basic engineering principles of system design which rely on identifying the most relevant feature of the domain for the problem at hand, and postulating an initial representation of the problem space organised around such a principal feature. Some features that have been favoured in the past include: causality, linear discourse, underlying structure, and character behavior. The present paper defends the need for simultaneous consideration of as many as possible of these aspects when attempting to model the process of creating narratives, together with some mechanism for distributing the weight of the decision processes across them. Humans faced with narrative construction may shift from views based on characters to views based on structure, then consider causality, and later also take into account the shape of discourse. This behavior can be related to the process of representational re-description of constraints as described in existing literature on cognitive models of the writing task. The paper discusses how existing computational models of narrative construction address this phenomenon, and argues for a computational model of narrative explicitly based on multiple aspects.