ipfs_eu_caselaw: Court of Justice of the European Union, English judgments (pack 1)

Target repo: endomorphosis/ipfs_eu_caselaw. Licence label: EU reuse (2011/833/EU; CURIA statement).

table	file	rows	source
cases	data/cases-00000.parquet	199	EU Publications Office Cellar (https://publications.europa.eu/resource/cellar/)

Selection: the 200 most recent Court of Justice judgments (CELEX 6YYYYCJNNNN; summaries, opinions and orders excluded) that have an English xhtml version, by cdm:work_date_document. The SPARQL query was run on 2026-10-29 at 01:44 PT; the query and its SHA256 are in scrapers/cellar_selection.json. Decision dates: 2025-11-13 .. 2026-10-15. One record was skipped by the family-law screen (maintenance obligations, Regulation 4/2009), leaving 199 rows. This is a sample, not the whole court: Cellar lists about 18,036 English CJ judgments with text.

Columns (same names as the NL caselaw packs, all strings): case_id (ECLI), ecli, jurisdiction (EU), court, court_id (eu-cjeu), document_type (judgment), decision_date, issued (empty), docket_numbers (JSON: CELEX and case numbers), title (official Cellar title; # separators shown as |), subject/procedure/judges/inhoudsindicatie (empty), language (en), parties ('' ; see parties_note), text, source_url (EUR-Lex CELEX page), api_url (Cellar https document URL), source_retrieved_at (UTC), source_sha256 (raw xhtml), provenance (JSON: endpoints, SPARQL query sha256, fetch time, collector/normalizer sha256, parser_version, CURIA case-list URL), license, license_text.

Text extraction: the xhtml body is walked once. Each table row becomes one line (the Court's layout puts each paragraph number and its text in adjacent cells), and each paragraph or heading outside a table becomes one line. Nothing is emitted twice and whitespace is normalised. Footnotes and signatures are kept. Text completeness against the raw xhtml body (non-whitespace characters) is 1.0 for every row.

Licence

EU reuse (2011/833/EU; CURIA statement). Basis, quoted verbatim:

1. CURIA legal notice (https://curia.europa.eu/common/juris/en/avertissement.html): "The texts available on this site may be reproduced provided the source is acknowledged as well as the fact that they are not authentic and are free of charge."
2. Publications Office copyright notice (https://op.europa.eu/en/web/about-us/legal-notices/publications-office-of-the-european-union-copyright, fetched 2026-10-29 01:34 PT): "The editorial content of this website, which is owned by the EU, is licensed under the Creative Commons Attribution 4.0 International licence (external link) . This means that you can reuse it provided you acknowledge the source and indicate any changes you have made."
3. Commission Decision 2011/833/EU, Art. 6(1) (via https://publications.europa.eu/resource/celex/32011D0833, fetched 2026-10-29 01:40 PT): "Documents shall be made available for reuse without application unless otherwise specified and without restrictions or, where appropriate, an open licence or disclaimer setting out conditions explaining the rights of reusers." Note: this Decision governs reuse of Commission documents; it is cited for context. The Court's own CURIA statement is the basis for the judgment texts.

Source acknowledged: Court of Justice of the European Union (curia.europa.eu), obtained through the EU Publications Office (Cellar). These texts are not authentic. Only the versions published in the European Court Reports and the Official Journal of the European Union are authentic. This dataset is provided free of charge.

Provenance and robots

• 1 SPARQL request (https://publications.europa.eu/webapi/rdf/sparql) and 200 document requests, 2026-10-29 01:44 PT - 2026-10-29 01:49 PT, 0 failures, at least 1 s between requests, one connection, User-Agent world-law-caselaw-research/0.1 (endomorphosis/ipfs_eu_caselaw; polite 1 req/s).
• Robots: publications.europa.eu redirects to https://op.europa.eu/robots.txt ("User-agent: * / Allow: /"; disallows only *requestAction=rdf, /webdav/*, portal/auth pages). The paths used were not disallowed. curia.europa.eu and eur-lex.europa.eu were not crawled.
• Builders are in scrapers/ (hashes in code.json); counts are in manifests/MANIFEST.md.

Personal data

The Court anonymises parties where required, and that anonymisation is kept as-is; no re-identification was attempted. PII sweep (national ID numbers, IBANs, Luhn-valid card numbers, personal webmail): 0 values redacted. Family-law matters (parental responsibility/child abduction/Brussels II, maintenance/matrimonial property/child adoption) and juvenile-justice matters (Directive 2016/800) are skipped when the official title names them.

Disclaimer

This dataset is provided for research and is not legal advice. It is not published by or affiliated with the Court of Justice of the European Union or the Publications Office. The texts may contain errors introduced by conversion; consult the authentic versions.

Takedown

If you are a party named in a decision or a rights holder and want a record removed or corrected, open a discussion on the Hugging Face repository endomorphosis/ipfs_eu_caselaw with the ECLI or CELEX number. The record will be removed from the next revision.