edithatogo/corpus-legislation-nz

New Zealand Legislation Corpus

Summary

This dataset is intended to contain machine-readable New Zealand legislation records collected through the official New Zealand Legislation API and published as optimized Parquet shards for live use. Coverage is not yet proven complete.

This dataset is part of the NZ corpus-family roadmap. The preferred systematic family label is corpus-nz-legislation, while this published Hugging Face dataset intentionally remains edithatogo/corpus-legislation-nz. The sibling Hansard corpus is corpus-nz-hansard.

Source provenance

Records include source URLs, API URLs, format URLs, hashes, scrape timestamps, and pipeline version metadata.

Data fields

Core fields are documented in docs/data_dictionary.md. They include:

• stable_id
• work_id
• version_id
• title
• legislation_type
• legislation_status
• source_url
• api_url
• xml_url
• html_url
• pdf_url
• text
• text_sha256
• source_hash

Intended use

• legal information retrieval
• statutory text search
• legal NLP experiments
• reproducible public-law data analysis
• annual comparison of statutory change

Limitations

• Corpus completeness is not yet proven. The current API-first pipeline depends on the configured discovery strategy, and search-based discovery must be reconciled against an authoritative inventory before any full-coverage claim.
• Text extraction is intentionally conservative and may not preserve all legal structure.
• Incorporated-by-reference material, third-party material, agency website text, logos, emblems, and non-legislative linked content may not be included or may have different rights.
• This dataset is not legal advice.

Loading examples

from datasets import load_dataset

ds = load_dataset("edithatogo/corpus-legislation-nz", split="train", streaming=True)
for row in ds.take(1):
    print(row["title"])

import pyarrow.dataset as ds

dataset = ds.dataset("data/parquet", format="parquet", partitioning="hive")
table = dataset.to_table(columns=["title", "legislation_type", "text"])

-- DuckDB example
SELECT title, legislation_type
FROM read_parquet('hf://datasets/edithatogo/corpus-legislation-nz/parquet/**/*.parquet')
LIMIT 10;

Update cadence

The live Hugging Face dataset is intended for daily or regular updates. DOI-backed snapshots are intended for annual Zenodo archival releases.

Citation

For academic or fixed-version citation, cite the Zenodo snapshot DOI: 10.5281/zenodo.20592540. For live use, cite the Hugging Face repository, access date, and the manifest hash from manifests/latest_manifest.json.

Licensing and legal caveat

The code that builds this dataset is licensed separately in the source repository. This dataset card does not relicense legislation text or third-party source material. Check the official New Zealand Legislation copyright page and your own institution's policies before redistributing or relying on the dataset. Zenodo archive metadata may use cc-by-4.0 for project-created metadata, packaging, manifests, checksums, and permitted source material; it must not be read as a blanket relicense of all upstream source material. This dataset is provided for information and research support, not legal advice.