Semantic Storytelling with Generative AI: An Ontology Driven Framework for Reliable Content Generation

In recent years, generative artificial intelligence - and in particular Large Language Models (LLM) - has revolutionised the way we interact with texts, giving rise to systems capable of generating complex content in natural language. However, the adoption of these tools in areas of high semantic sensitivity, such as cultural valorisation, education, institutional communication or regulatory analysis, raises critical questions related to source reliability, narrative coherence and informational correctness. This thesis proposes an innovative methodology that integrates computational storytelling, semantic representation through ontologies and Retrieval-Augmented Generation (RAG) techniques, with the aim of guiding and controlling the generative activity of LLMs. The developed model enables the structuring of stories in a formal, verifiable and machine-interpretable manner, supporting the automatic generation of coherent, contextualised and source-accurate narratives. The methodology was applied and validated in six real use cases, in collaboration with public administrations and cultural institutions, including the Court of Naples, the Central State Archive of Rome, the IIIAC-CSIC in Barcelona and the Municipality of Castellammare di Stabia. In several of these application cases, Augmented Reality (AR) and Virtual Reality (VR) user scenarios have been developed, allowing for conversational interaction with avatar agents within immersive narrative experiences. The cases range from the automatic narration of historical places, education through immersive narrative agents, communication in emergency contexts, to the comparative analysis of legal texts and semantic navigation in transmedia environments. The results demonstrate the effectiveness of the proposed model in reducing the hallucinations of language models, improving the semantic quality of responses and promoting an ethical, controllable and transparent use of artificial intelligence in narrative and communicative processes in public administration.