LexOnt-IT & LexQuery: An OWL 2 DL Ontology and a Natural-Language Interface for Italian Case Law

This paper presents LexOnt-IT, an OWL 2 DL ontology for the formal semantic representation of the case law of the Italian Supreme Court of Cassation, and LexQuery, a natural-language querying engine that exploits the ontological schema as a grounding foundation and large language models (LLMs) to enable conversational access to any OWL-based legal knowledge graph. LexOnt-IT models decisions, legal institutes, statutory provisions, judicial panels, and inter-decision citation networks, aligning with the ELI, ECLI, Akoma Ntoso, and NIR international standards. A pipeline called K-Extractor automatically populates the knowledge graph from PDF judgments via NERand RDF triple generation. LexQuery exploits ontology schema grounding and retrieval-augmented generation (RAG) to translate natural language questions into validated SPARQL queries, enabling non-expert users—magistrates, practitioners, and researchers—to query the Italgiure knowledge graph without SPARQL expertise. We describe the system architecture, ontology design choices, pipeline implementation, and a two-tier benchmark evaluation of 20 queries, reporting RF1=0.950 on the five independently formulated queries and overall RF1=0.908. An ablation study confirms that schema grounding raises RF1 by 54.1 percentage points over an ungrounded baseline. We further discuss the possibility of the portability of LexQuery to EUR-Lex CELLAR and future extensions towards a pan-European interface for legal knowledge graphs.