Search papers, labs, and topics across Lattice.
This paper introduces a two-stage LLM-assisted workflow designed to enhance the extraction and structuring of French maintenance regulations into a legal knowledge graph. By utilizing a SEMLEG-based core ontology, the authors successfully extract typed entities and triples from a stratified corpus, achieving near-complete class alignment and significant reductions in duplicated entities. The findings highlight the importance of predicate normalization and the validation of new relation signatures, which are crucial for improving the utility of legal texts in operational systems.
The integration of LLMs in ontology engineering reveals that fewer than 20% of extracted triples introduce unseen properties, challenging assumptions about legal data completeness.
Maintenance regulations are complex legal texts that are difficult to exploit when addressing a specific case and challenging to integrate into operational systems. This paper presents a two-stage LLM-assisted workflow for French maintenance regulations: ontology engineering from a SEMLEG-based core ontology, followed by construction of an ontology-grounded French legal knowledge graph. The first stage consists in the open extraction of typed entities and triples from a stratified corpus sample, the normalization of labels through embedding-based fusion, and the induction of candidate object properties with their signature (domain and range). The second stage uses the resulting ontology to guide the closed extraction of triples and RDF graph construction over the full corpus. Experiments with GPT-4.1 and mistral-large-2512 show robust structured outputs, near-complete class alignment, and a substantial reduction of duplicated entities and predicates after fusion. Fewer than 20% of triples introduce unseen properties, while lower exact signature compliance reveals new domain-range combinations for existing predicates. These results point to predicate normalization and the validation of newly observed relation signatures as key refinement steps for industrial maintenance settings.
