JUDDGES

JuDDGES

Judicial Decision Data Gathering, Encoding and Sharing — an Open Science platform for scalable, transparent, human-validated judicial decision analysis across jurisdictions.

NLP + HITL Open Data & Software Cross-jurisdiction Audit-ready exports
Middlesex University
Lyon Partner University
Poland Partner University

Powered by Margin-Aware Active Learning

The system can prioritise the most informative decisions for expert review (e.g., ambiguous/low-confidence cases), improving consistency and reducing annotation cost over time.

Project Overview

JuDDGES harnesses state-of-the-art Natural Language Processing and Human-in-the-Loop (HITL) methods to enable scalable, flexible, and on-going meta-annotation — combining automated suggestions with domain experts in-the-loop.

The platform supports legal records and judgments across jurisdictions, including England & Wales, Poland, and partner work in Lyon (France).

Use-case: criminal judgments Method: expert-validated annotation Outcome: reusable datasets

Why this matters

We reduce barriers of resources, language, and data/format inhomogeneity that impede research on judicial decision-making. Open tools + open outputs enable researchers to empirically test theories and inform evidence-led policy questions.

Designed for legal researchers — simple interface, research-grade traceability.

How the platform works (AI + Human-in-the-Loop)

Workflow

  1. Ingest legal decisions into a searchable corpus
  2. Suggest topics/codes using NLP models
  3. Validate with expert review (accept / reject / correct)
  4. Audit decisions and maintain traceability
  5. Export structured datasets for analysis (CSV/SPSS-ready)
AI accelerates discovery; human expertise protects validity and interpretability.

What you can do

Search at scale Meta-annotate Compare jurisdictions Track decisions Reuse & share

Open resources

Zenodo Community

Open outputs: coding schemes, reports, workshop materials, and publications.

Datasets (HuggingFace)

Project datasets and structured releases: https://huggingface.co/JuDDGES

Selected publications

Abbas, W., Zia, T., Tirunagari, S., Chennareddy, V., Dhami, M., & Windridge, D. (2026). The Application of Pre-trained Transformer Models to UK Court of Appeal Legal Judgments. Springer Nature. DOI

Dhami, M., Kajdanowicz, T., Windridge, D., & Boukacem-Zeghmouri, C. (2025). Judges on Trial: The Future of Research on Judicial Decision-Making. Wiley. DOI

Tirunagari, S., Dhami, M., & Windridge, D. (2025). Margin-aware Active Learning for User-adaptive Text Classification. PDF

Funding, Acknowledgements & Contact

Project funded by:
Engineering and Physical Sciences Research Council (EPSRC)
Grant EP/Y035992/1Judicial Decision Data Gathering, Encoding and Sharing

Supported under the CHIST-ERA Call ORD “Open & Re-usable Research Data & Software”

Contact:
Prof. Mandeep K. Dhami


Prof. David Windridge

UKRI EPSRC CHIST-ERA
Acknowledgements
We are grateful to Charis Bechan and Vili Hadzhieva for their annotation contributions.

We also thank Mahensingh Deonaran (Middlesex University) for curating the initial dataset version and supporting early data preparation.
Open Science Open Infrastructure Public Reuse
If you are a public institution and want to reuse the data/tooling, please get in touch for access and guidance.