From Judgments to Issues: Structured Extraction of Legal Reasoning with Citation-Hallucination Control
arXiv:2607.03325v1 Announce Type: cross Abstract: We present an automated pipeline that decomposes Italian tax-court judgments into individual legal issues and extracts, for each issue, a structured XML representation grounded in the IRAC framework and the legal syllogism. The pipeline targets a corpus of approximately $330{,}000$ first- and second-instance decisions of the Italian tax courts and is built around a capable yet cost-efficient general-purpose model (DeepSeek V3), a choice driven by the need to process several hundred thousand documents at a sustainable cost. To address the well-d
The proliferation of capable and cost-efficient general-purpose AI models, exemplified by DeepSeek V3, is enabling the automation of complex, domain-specific tasks that were previously cost-prohibitive.
This development indicates a tangible path towards applying AI to structure vast quantities of unstructured legal data, which has significant implications for legal efficiency and access to justice.
The ability to automatically decompose legal judgments into structured, Issue-Rule-Application-Conclusion (IRAC) based formats transforms how legal information can be processed, analyzed, and retrieved.
- · Legal Tech companies
- · Legal services sector
- · Governments/Judiciaries
- · AI model developers
- · Traditional legal research services
- · Data entry/Paralegal services
- · Jurisdictions with unstructured legal data
Automated legal reasoning extraction significantly reduces the time and cost associated with legal research and case analysis.
Improved access to structured legal precedents could lead to more consistent judicial decision-making and potentially accelerate legal processes.
The structured legal data could form the basis for advanced legal expert systems, potentially challenging established legal frameworks and professions.
This signal links to a primary source. Continuum Brief monitors and indexes it as part of the live intelligence stream — we do not republish source content.
Read at arXiv cs.AI