An AI-powered approach to the semiotic reconstruction of narratives

Abstract

This article presents a novel and highly interactive process to generate natural language narratives based on our ongoing work on semiotic relations, providing four criteria for composing new narratives from existing stories. The wide applicability of this semiotic reconstruction process is suggested by a reputed literary scholar’s deconstructive claim that new narratives can often be shown to be a tissue of previous narratives. Along, respectively, three semiotic axes – syntagmatic, paradigmatic, and meronymic – existing stories can yield new stories by the combination, imitation, or expansion of an iconic scene; lastly, a new story may emerge through reversal via an antithetic consideration, i.e., through the adoption of opposite values. Targeting casual users, we present a fully operational prototype with a simple and user-friendly interface that incorporates an AI agent, namely ChatGPT. The prototype, in a coauthor capacity, generates context-compatible sequences of events in storyboard format using backward-chaining abductive reasoning (employing Stable Diffusion to draw scene illustrations), conforming as much as possible to the user’s authorial instructions. The extensive repertoire of book and movie summaries available to the AI agent obviates the need to manually supply laborious and error-prone context specifications. A user study was conducted to evaluate user experience and satisfaction with the generated narratives. The preliminary findings suggest that our approach has the potential to enhance story quality while offering a positive user experience.