Revisiting Computational Models of Creative Storytelling Based on Imaginative Recall

Abstract

Certain story generation systems consider the processes of imaginative recall and adaptation as central to human creativity in storytelling. Researchers have recently compared the output of these systems through the lens of Boden’s types of creativity [9]. This comparison highlights the contribution of predefined structures to story predictability, which influences perceived creativity. We revisit the connection between knowledge structures and story predictability, and compare Minstrel’s use of knowledge structures versus the use of Story Intention Graphs (SIGs) as the underlying case frames. Semantic information encoded in the SIG produces coherent stories and retains the imaginative recall and generalization aspect of Minstrel’s creative process. Mapping knowledge structures to SIGs enables the use of a common representation that is directly connected to surface realization. This opens up the performative aspect of creativity that does not come out in templated text outputs.