chenjigaram/eu-financial-regulation-aligned

EU Financial Regulation, Aligned Across 24 Languages

Eight EU financial regulations, split to the paragraph, in every official EU language, with the alignment verified rather than assumed.

56,838 rows — 2,442 provisions × up to 24 languages.

Why the alignment is exact

Most multilingual legal corpora are aligned by matching sentences, which is approximate and fails on exactly the long provisions people care about.

This one does not do that. EUR-Lex assigns structural ids per act, not per translation: Article 16 is art_16 in every language version, and its seventh paragraph is 016.007 whether you asked for English, Croatian or Greek. Rows sharing a unit_id are the same provision by construction.

That claim was checked across all eight acts in all 24 languages — 181 act-language pairs:

	Pairs
Article ids identical to English	181/181	100%
Article and paragraph ids identical	179/181	98.9%

Contents

Act	CELEX	Articles	Provisions
MiFID II — Directive 2014/65/EU	32014L0065	97	406
MiFIR — Regulation (EU) No 600/2014	32014R0600	55	231
PRIIPs — Regulation (EU) No 1286/2014	32014R1286	34	89
UCITS — Directive 2009/65/EC	32009L0065	119	319
AIFMD — Directive 2011/61/EU	32011L0061	71	363
DORA — Regulation (EU) 2022/2554	32022R2554	64	254
MiCA — Regulation (EU) 2023/1114	32023R1114	149	732
SFDR — Regulation (EU) 2019/2088	32019R2088	20	48

Schema

Field	Meaning
unit_id	CELEX:article:paragraph — the join key across languages
celex	CELEX identifier of the act
act, act_name, domain	short key, display name, subject area
article, paragraph, anchor	position within the act
language	ISO 639-1 code
heading	article title, or the chapter/section chain where the act gives none
text	the provision
url	deep link to the article on EUR-Lex

Building translation pairs

from collections import defaultdict
from datasets import load_dataset

rows = load_dataset("chenjigaram/eu-financial-regulation-aligned", split="train")

units = defaultdict(dict)
for row in rows:
    units[row["unit_id"]][row["language"]] = row["text"]

pairs = [(u["en"], u["nl"]) for u in units.values() if "en" in u and "nl" in u]

Any of the 552 ordered language pairs works the same way.

Coverage and its gaps

Languages per provision	Provisions
24	1,354
23	406
22	682

Every provision is present in at least 22 languages. The shortfall is three whole acts missing a language, not scattered holes:

• No Irish for MiFID II, UCITS and AIFMD. The Irish language derogation ran until 2022, so acts adopted before then were never published in Irish.
• No Croatian paragraphs for UCITS and AIFMD. Croatia joined in 2013; earlier acts appear in Croatian as a special edition in the older EUR-Lex layout, which carries no structural ids. Those two acts are present in Croatian at article level but cannot be aligned below it, so their paragraph rows are absent.

Known limitations

• Article bodies only. Recitals, annexes and tables are not included.
• Original adopted texts, not consolidated versions — amendments since adoption are not reflected.
• heading is empty for PRIIPs and UCITS articles, because those acts give their articles no titles; the chapter and section chain is used instead.

Licence and attribution

Source: EUR-Lex, © European Union. Reused under Commission Decision 2011/833/EU; editorial content is licensed CC BY 4.0.

Only the authentic printed Official Journal has legal value. Do not rely on this for anything where that distinction matters.

Reproducing it

Built with eurlex-corpus:

pip install eurlex-corpus
python tools/build_dataset.py
python tools/check_alignment.py