Integrating Wikidata Entities into Narrative Graphs Using Large Language Models

Narratives are essential tools for articulating and sharing human experiences, particularly in scientific and cultural domains where they aid in the explanation of complex phenomena. In the context of a broader scientific effort in which Knowledge Representation and Semantic Web technologies are used to transform raw textual data into formal narratives, this paper explores the capability of Large Language Models (LLMs) to associate narrative entities (e.g. persons, locations, keywords) with the corresponding entity identifiers from Wikidata. We propose three LLM-based approaches for extracting and linking narrative entities and compare their performance against the JSI Wikifier, a state of-the-art entity linking tool. The evaluation is based on a dataset from the H2020 MOVING project, which focuses on sustainability and value chains in European mountain regions. This study highlights the potential of LLMs to improve semantic annotation workflows and contribute to the automated generation of semantically enriched narratives.