Neurosymbolic Narrative Generation for Cultural Heritage

Abstract

Aim of my research is to exploit Linguistic Linked Open Data (LLOD) as base for advanced Cultural Heritage (CH) fruition by means of Automatic Story Generation (ASG). Following the rationale that discovering and reviving already existing (yet latent) narratives is worthier than automatically generating them from anew in eliciting the user's interest, the input-2-graph and the graph-2-sequence ASG-pipeline phases, heavily relying on LLOD, will be given a deeper focus, whereby the final Natural Language Generation (NLG) module will be constrained by the entities and relations established in the Knowledge Graph (KG) generation modules (a configuration typical of the neurosymbolic approach). In order to enhance possibilities of implementation in real-life contexts, the elaborated pipeline will be modular, i.e. self-sufficient in its constituent parts. Beyond the countless possible application scenarios ranging from education to entertainment, this solution detangles the user from his role of mere consumer, and empowers him not only to control the creation process [3.1], but also to find already within it, and not necessarily in the final outcome, a valuable source for intellectual growth. This work intends the addressing of a specific societal need as an avalanche to simultaneously fill knowledge gaps identified in and among the related scientific domains.