Legal Requirements Translation from Law

Fuente: Zenodo
Saved in:
Bibliographic Details
Main Authors: Singhal, Anmol, Breaux, Travis
Format: Recurso digital
Published: Zenodo 2025
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866902296138350592
author Singhal, Anmol
Breaux, Travis
author_facet Singhal, Anmol
Breaux, Travis
contents <p>This artifact provides a framework for translating legal requirements from natural language into structured, executable Python code using GPT-4o. The representation captures not only individual metadata elements but also their semantic and structural relationships—such as obligations, conditions, and exceptions—across legal clauses. This enables downstream tasks such as compliance verification.</p> <p>The artifact includes:</p> <ul> <li> <p>A domain-agnostic class structure for modeling legal requirements.</p> </li> <li> <p>Prompt templates and translation methodology using GPT-4o.</p> </li> <li> <p>Testing notebooks for evaluating structural accuracy, semantic correctness, and code compilability.</p> </li> <li> <p>Evaluation metrics including <code>pass@k</code> and attribute-level precision/recall.</p> </li> <li> <p>Test datasets from U.S. state data breach notification laws and sample model outputs.</p> </li> </ul> <p>The approach is designed to be extensible to other legal domains or jurisdictions due to the flexibility and generality of the class design. Researchers and practitioners can adapt the translation pipeline to other LLMs and legal texts.</p> <p>For installation instructions, usage examples, and system requirements, please refer to the README file. Intermediate outputs are also provided for reproducibility and quick validation.</p>
format Recurso digital
id zenodo_https___doi_org_10_5281_zenodo_15794182
institution Zenodo
language
publishDate 2025
publisher Zenodo
record_format zenodo
spellingShingle Legal Requirements Translation from Law
Singhal, Anmol
Breaux, Travis
<p>This artifact provides a framework for translating legal requirements from natural language into structured, executable Python code using GPT-4o. The representation captures not only individual metadata elements but also their semantic and structural relationships—such as obligations, conditions, and exceptions—across legal clauses. This enables downstream tasks such as compliance verification.</p> <p>The artifact includes:</p> <ul> <li> <p>A domain-agnostic class structure for modeling legal requirements.</p> </li> <li> <p>Prompt templates and translation methodology using GPT-4o.</p> </li> <li> <p>Testing notebooks for evaluating structural accuracy, semantic correctness, and code compilability.</p> </li> <li> <p>Evaluation metrics including <code>pass@k</code> and attribute-level precision/recall.</p> </li> <li> <p>Test datasets from U.S. state data breach notification laws and sample model outputs.</p> </li> </ul> <p>The approach is designed to be extensible to other legal domains or jurisdictions due to the flexibility and generality of the class design. Researchers and practitioners can adapt the translation pipeline to other LLMs and legal texts.</p> <p>For installation instructions, usage examples, and system requirements, please refer to the README file. Intermediate outputs are also provided for reproducibility and quick validation.</p>
title Legal Requirements Translation from Law
url https://doi.org/10.5281/zenodo.15794182