LOCUS-v1

# LOCUS v1.0 This repository contains the dataset presented in the paper [Freeing the Law with LOCUS: A Local Ordinance Corpus for the United States](https://huggingface.co/papers/2606.19334). ## Dataset Summary LOCUS v1.0 is a chunk-level dataset of U.S. municipal and county law text labeled by legal function. Each eligible chunk is assigned a `function`, a binary `is_substantive` label, and all substantive provisions are assigned a `topic`. The dataset is intended for legal text research, local-law structure analysis, substantive filtering, and downstream taxonomy refinement. ## Dataset Structure The unit of analysis is a text chunk derived from local law documents. ## Scale Approximate scope for v1: `2,211,516` ordinances ## Label Schema ### Function Allowed values: - `Context` - `Rules` - `Process` - `Enforcement` ### Substantive Indicator The released dataset enforces the following deterministic rule: - `is_substantive = 1` for `Rules` and `Enforcement` - `is_substantive = 0` for `Context`, `Process`, and `Structural` ### Topic Used only when `is_substantive = 1`. Allowed values: - `Buildings` - `Business` - `Nuisance` - `Zoning` - `Other` ## Recommended Uses LOCUS v1.0 is appropriate for: - legal text classification research - local law structure analysis - substantive versus non-substantive filtering - downstream taxonomy refinement - weakly supervised or human-in-the-loop legal NLP workflows ## Out-of-Scope Uses LOCUS v1.0 should not be treated as: - legal advice - a substitute for human legal review - a complete census of all U.S. local law - a fully human-validated benchmark without additional auditing ## Citation If you use this dataset, please cite: ```bibtex @article{peskoff2026freeing, title={Freeing the Law with LOCUS: A Local Ordinance Corpus for the United States}, author={Peskoff, Denis and Barrow, Joe and Vu, Christopher and Davenport, Diag}, journal={arXiv preprint arXiv:2606.19334}, year={2026} } ```