Analysis of the correlations between the knowledge structures of an automatic storyteller and its literary production

Abstract

In this paper, we describe a process to identify relations among the features of the knowledge base of an automatic storyteller and the narratives that it generates. We define structures to analyze the internal composition of the information available for an agent. We also establish a set of metrics to identify diverse story characteristics. Next, we perform experiments utilizing Mexica, an automatic storyteller, to generate narratives and to evaluate them according to our set of metrics. Then, we compare such assessments with the visual structures that we built from the agent’s original knowledge base, in order to obtain correlations between them. The results suggest that such correlations are useful to study the links between the agents’ knowledge base and the kind of stories they might produce.