| _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 |