Romanian Law Data

Romanian legal corpus (documents, articles, paragraphs, relationships) as zstd parquets for DuckDB.

From legislatie.just.ro (Ministry of Justice), scraped via the public HTML.

Rebuilt daily by GitHub Actions. Download from Releases.

Tables

Table	Content	Rows
documents	One row per act (LEGE, OUG, HG, ORDIN, DECIZIE, ...) with metadata + full text	~187k
articles	One row per article, parsed from documents.content	~993k
paragraphs	One row per alineat, the finest citation unit (e.g. art. 188 alin. (1))	~1.96M
relationships	One row per directed edge (act → target): repeals, amends, suspends, ...	varies

English column names (type, document_number, published_at, gazette_number, ...) with Romanian COMMENT ON metadata in create_views.sql.

Usage

Download latest release, and open it using DuckDB:

mkdir -p data
gh release download -R scriptogre/romanian-law-data --pattern '*.parquet' --pattern 'create_views.sql' --dir data

import duckdb
conn = duckdb.connect()
conn.execute(open("data/create_views.sql").read())
conn.execute("""
    SELECT document_citation, link, article_citation, content
    FROM articles
    WHERE document_id IN (SELECT id FROM penal_code)
      AND article_number = 188
""").fetchall()