Knowledge Graphs Generation from Cultural Heritage Texts: Combining LLMs and Ontological Engineering for Scholarly Debates

TL;DR

This work addresses the difficulty of converting unstructured Cultural Heritage discourse into queryable Knowledge Graphs by introducing ATR4CH, a five-step, LLM-centric methodology that tightly couples annotation, ontological alignment, and knowledge extraction. The SEBI ontology and RDF-star-based representation enable modeling of multi-perspective scholarly claims, evidence, and hypotheses within CH debates. Empirical results across metadata, entity recognition, evidence, and hypothesis tasks show high performance for metadata ($F_1$ near $0.99$), strong evidence extraction, and competitive entity and hypothesis extraction, with smaller models offering cost-effective deployment. Limitations include reliance on Wikipedia as the data source and the need for human oversight; future work targets multilinguality, primary sources, and broader CH domain applicability, expanding to multilingual and primary-literature contexts with enhanced human-in-the-loop tooling.

Abstract

Cultural Heritage texts contain rich knowledge that is difficult to query systematically due to the challenges of converting unstructured discourse into structured Knowledge Graphs (KGs). This paper introduces ATR4CH (Adaptive Text-to-RDF for Cultural Heritage), a systematic five-step methodology for Large Language Model-based Knowledge Extraction from Cultural Heritage documents. We validate the methodology through a case study on authenticity assessment debates. Methodology - ATR4CH combines annotation models, ontological frameworks, and LLM-based extraction through iterative development: foundational analysis, annotation schema development, pipeline architecture, integration refinement, and comprehensive evaluation. We demonstrate the approach using Wikipedia articles about disputed items (documents, artifacts...), implementing a sequential pipeline with three LLMs (Claude Sonnet 3.7, Llama 3.3 70B, GPT-4o-mini). Findings - The methodology successfully extracts complex Cultural Heritage knowledge: 0.96-0.99 F1 for metadata extraction, 0.7-0.8 F1 for entity recognition, 0.65-0.75 F1 for hypothesis extraction, 0.95-0.97 for evidence extraction, and 0.62 G-EVAL for discourse representation. Smaller models performed competitively, enabling cost-effective deployment. Originality - This is the first systematic methodology for coordinating LLM-based extraction with Cultural Heritage ontologies. ATR4CH provides a replicable framework adaptable across CH domains and institutional resources. Research Limitations - The produced KG is limited to Wikipedia articles. While the results are encouraging, human oversight is necessary during post-processing. Practical Implications - ATR4CH enables Cultural Heritage institutions to systematically convert textual knowledge into queryable KGs, supporting automated metadata enrichment and knowledge discovery.

Figures (23)

• Figure 1: The Donation of Constantine entry in Wikidata
• Figure 2: Flowchart of the ATR4CH methodology showing the five-step iterative process
• Figure 3: Overall distribution of article lengths showing the right-skewed pattern characteristic of encyclopedic content, with most articles in the 2k-15k character range and notable outliers extending beyond 40k characters.
• Figure 4: Distribution of articles across Wikipedia categories, showing the natural prevalence of different forgery types in scholarly discourse.
• Figure 5: Token count distribution by category, illustrating variability in article length and content density. Box plots show medians, quartiles, and outliers representing comprehensive case studies.