endomorphosis/ipfs_nigeria_laws

Nigeria Constitution and Acts of the National Assembly (official hosts)

Research snapshot of official national legislation from Federal Ministry of Justice + National Assembly official hosts.

Not legal advice. The official gazette / authentic source prevails over this corpus.

Snapshot

Field	Value
Snapshot date	2026-09-10 (base) + 2026-10-03 (NG national EXPAND, live gov.ng)
Coverage	official-hosted-snapshot
Source	Federal Ministry of Justice + National Assembly official hosts
Collector	scrapers/collect_ng.py (base), scrapers/ng_collect.py (expand)
Laws / instruments	394
Articles	11,662
Language	en
Jurisdiction	Nigeria
License	ng-official-texts

Contents

• data/laws.parquet — one row per instrument.
• data/articles.parquet — article/section rows for the same snapshot.
• scrapers/collect_ng.py — research collector used to build the base snapshot (222 instruments).
• scrapers/ng_collect.py — collector for the 2026-10-03 national EXPAND (live gov.ng only).

Expansion 2026-10-03 — national EXPAND (append-only)

• Appended: 140 instruments / 3,805 rows to the 222 / 7,216 base rows. The base rows are unchanged (never-thin).
• Source: live official government hosts only.
  • 139 Acts come from the National Assembly "Acts & Publications" catalog (https://nass.gov.ng/documents/magazine, downloads https://nass.gov.ng/documents/download/<id>).
  • 1 instrument (Child Rights Enforcement Procedure Rules 2015) comes from justice.gov.ng.
  • Fetched 2026-10-03 PT. robots.txt was obeyed, with at least 2.5 s between requests per host. No LII, mirror or web-archive copies were used.
• Dedup: against the base snapshot by NASS download id, normalized title + year / Cap number, and text-shingle similarity. 6 catalog items were left out: 3 identical re-uploads, 2 already in the base snapshot, and 1 whose official PDF is corrupt.
• Text extraction: pdftotext word boxes are used, with marginal notes separated from the body text.
  • Where the PDF text layer is missing or poor, pages are OCR'd with Tesseract (tesseract 5.5.0 eng, pdftoppm 300dpi gray, psm 3, English, 300 dpi). This applied to 64 documents (1,231 pages).
  • Each instrument records its method (text_layer / text_layer+ocr_pages / ocr) and quality metrics in metadata_json.metadata.text_extraction: dictionary-word ratio, garbage-character ratio, characters per page, and grade.
  • Method counts: {'text_layer': 76, 'text_layer+ocr_pages': 13, 'ocr': 51}. Median dictionary-word ratio 0.956.
  • 4 instruments fall below the quality threshold (dict ≥ 0.85, garbage ≤ 0.02) and are flagged.
• Rows:
  • One row per numbered section (<id>-section-<n>). Court rules are split by Order.
  • The text before section 1 (title, explanatory memorandum, arrangement, long title) becomes a preamble annex row, and schedules become schedule annex rows.
  • Some scans lack reliable section numbers. In that case the complete text is kept as ONE text row and flagged whole_text_fallback.
  • Instruments where fewer sections were recovered than the document's own numbering are flagged partial_sections. In those, all the text is still present: unsplit sections are merged into their neighbours.
  • Section split status: {'OK': 82, 'PARTIAL': 34, 'WHOLE_TEXT_FALLBACK': 24}.
• Provenance note on the base snapshot: 175 of the 222 base instruments were retrieved from web-archive (Wayback id_) copies of official nass.gov.ng / nigeriarights.gov.ng URLs. See metadata_json.metadata.retrieval; source_url shows the official URL. These rows are flagged and left unchanged.

License / reuse

Official texts of the Federal Republic of Nigeria; Official Gazette prevails; not legal advice.

This dataset is not legal advice. The official source prevails.

Source

• https://nass.gov.ng/
• https://justice.gov.ng/

Known gaps

Constitution (FMOJ PDF) + Acts hosted on nass.gov.ng/documents/download and Wayback of those official URLs. Not LawNigeria/LawPavilion. gazettes.africa not crawled (robots Disallow). NASS SPA lists only recently uploaded Acts; LFN 2004 consolidations are not a complete public catalog on justice.gov.ng.

Update 2026-10-10

Additive merge: 4 new instruments (37 new article rows) from the official-source local corpus (scrapers/cc_country_harvest.py (Common Crawl ranged WARC)); all 362 previously published instrument rows and 11021 article rows are carried verbatim (no removals; merge is by instrument id with URL/text de-duplication). Quality gate dropped non-legislative / garbled / duplicate local files (see MANIFEST). Instruments with no source article split were split on language-specific article headings (backend article_repair_aae).

Batch 2 update (2026-10-16, Africa & Middle East lane)

Additive update on top of endomorphosis/ipfs_nigeria_laws@5b1696af4325: laws/articles 366/11,058 -> 394/11,662 (394/11662). New instruments: 28 (+604 article rows). Every row that was already in this dataset is kept; only this lane's own ng-cc- rows may carry title or language repairs (title repairs kept: 0, language repairs: 0).

• Source: Common Crawl captures (index -> WARC records) of official government, parliament, gazette and regulator hosts for this country; no live requests. Court and news sites are excluded.
• New rows must be a single law instrument. News and press items, speeches, reports and plans, forms, tenders and notices, bills and drafts, court rulings, web stubs, budget tables and whole gazette issues (over about 200k characters, or carrying several instruments) are rejected, and rows that duplicate a row already here (same text, or same URL ignoring tracking parameters) are dropped.
• The pack was audited before release (row-level checks against the base commit, title review). Rows the audit rejected were removed, bad titles were rewritten from the document text, and lane title changes judged wrong were reverted to the earlier title. Counts: dropped 36, retitled 6.
• Article rows are an exact partition of the law text at numbered article/section markers; laws with no clean markers have no article rows.
• Per-row provenance (source URL, crawl, dropped candidates and reasons, title/language repairs, audit fixes) is in provenance/changes.json; file hashes in MANIFEST.json.
• Research compilation of public legal texts. The official gazette / authentic text prevails. Not legal advice.