GELATO

This repo contains the data from "The Gelato Dataset for Legislative NER" (LREC2026).

Dataset Description

GELATO (Government, Executive, Legislative, and Treaty Ontology) is a dataset of U.S. House and Senate bills from the 118th Congress annotated using a novel two-level named entity recognition ontology designed for U.S. legislative texts.

• Language: English
• License: MIT

Dataset Sources

• Repository: GitHub
• Paper: The GELATO Dataset for Legislative NER

Uses

This dataset contains a two-level ontology to support NLP research of U.S. legislative data.

Dataset Structure and Ontology

1. Person
   1. Individual
   2. Member
   3. Title
2. Organization
   1. Agency
   2. Association
   3. Committee
   4. International Institution
   5. Legislative Body
   6. Locality
   7. Nation
   8. State
3. Document
   1. Bill
   2. Code
   3. Parenthetical
   4. Reference
   5. Report
   6. Treaty
4. Abstraction
   1. Case
   2. Doctrine
   3. Fund
   4. Infrastructure
   5. Misc
   6. Program
   7. Session
   8. Specification
   9. System
5. Act
   1. Amendment
   2. Public Act
6. Class
   1. Non-Protected Class
   2. Protected Class

Source Data, Data Collection, and Processing

All bills in the GELATO dataset are publicly available U.S. government documents obtained via the congress.gov API and are therefore in the public domain and not subject to copyright restrictions.

Annotation process and annotators

The three graduate student authors annotated the data following best practices. See our paper for more details.

Bias, Risks, and Limitations

Three graduate student annotators (the authors) with training in linguistics and NLP collaboratively created GELATO through a two-stage process with full adjudication of any disagreements. This is a descriptive annotation; for example, this ontology includes Protected Class and Non-Protected Class subclasses that are consistent with U.S. anti-discrimination law definitions.

GELATO can support beneficial applications including legislative tracking, policy analysis, and government transparency initiatives. However, automated entity extraction could also enable potentially harmful uses such as targeted analysis of how specific groups are referenced in legislation or identification of individual legislators for inappropriate purposes.

Citation

BibTeX:

@misc{flynn2026gelatodatasetlegislativener,
      title={The GELATO Dataset for Legislative NER}, 
      author={Matthew Flynn and Timothy Obiso and Sam Newman},
      year={2026},
      eprint={2603.14130},
      archivePrefix={arXiv},
      primaryClass={cs.CL},
      url={https://arxiv.org/abs/2603.14130}, 
}

Dataset Card Authors

Matthew Flynn (Wollaston)

Dataset Card Contact

Matthew Flynn (Wollaston)