Dataset Card for Vietnamese Legal Corpus (UTS_VLC)

A curated corpus of Vietnamese Laws and Codes (Luật, Bộ luật) and the Constitution, maintained by Underthesea NLP. The flagship 2026 split is a verified in-force snapshot — every document is currently in force, de-duplicated, and validated against Vietnam's official legal database vbpl.vn.

Dataset Details

Dataset Description

UTS_VLC contains the full text of Vietnamese legislation at the top of the legal hierarchy: the Constitution, the Codes (Bộ luật), and the Laws (Luật) passed by the National Assembly, from 1945 to 2026. Each document is provided as cleaned Markdown together with its official document number (Số hiệu) and type.

• Curated by: Underthesea NLP
• Last updated: 2026-06-01
• Language(s): Vietnamese (vi)
• License: MIT
• Repository: https://huggingface.co/datasets/undertheseanlp/UTS_VLC
• Primary source: vbpl.vn — Cơ sở dữ liệu quốc gia về pháp luật (Ministry of Justice)

Vietnamese Legal Hierarchy

The dataset encodes the Vietnamese legal document hierarchy (per Luật ban hành văn bản quy phạm pháp luật 2015):

Level	Document types	Issuing authority
1	Hiến pháp (Constitution)	Quốc hội
2	Luật, Bộ luật (Laws, Codes)	Quốc hội
3	Pháp lệnh, Lệnh (Ordinances, Orders)	UBTVQH / Chủ tịch nước
4	Nghị định, Nghị quyết (Decrees, Resolutions)	Chính phủ
5	Thông tư, Quyết định (Circulars, Decisions)	Bộ trưởng / UBND
6	Chỉ thị (Directives)	Thủ tướng / Chủ tịch UBND

UTS_VLC covers the top of this hierarchy — the Constitution and, primarily, Level 2 (Luật & Bộ luật) issued by the National Assembly.

Uses

Direct Use

• Language modelling / pre-training on Vietnamese legal text.
• Legal text classification (by domain, type, or status).
• Legal question answering and retrieval over current Vietnamese law.
• Fine-tuning and evaluation where clean, de-duplicated, in-force legal text is required.

from datasets import load_dataset

ds = load_dataset("undertheseanlp/UTS_VLC", split="2026")       # 306 docs (Constitution + 305 in-force Laws/Codes)
raw = load_dataset("undertheseanlp/UTS_VLC", split="2026_01")   # 318, original Jan 2026 (issue #1035)

codes = ds.filter(lambda x: x["type"] == "code")                # the 6 Codes (Bộ luật)
land_law = ds.filter(lambda x: "Đất đai" in x["title"])         # search by title

Out-of-Scope Use

• Not legal advice. Texts may contain amendments not yet consolidated; always verify against the official source for legal decisions.
• The 2026_01, 2023, and 2021 splits include superseded/expired documents and should not be treated as current law.
• The corpus stops at Level 2 of the hierarchy — it does not contain Decrees, Circulars, etc.

Dataset Structure

Splits (config default)

Split	Documents	Description
2021	110	Original corpus with full text
2023	208	Expanded corpus with processed text
2026_01	318	Original 2026 release (Jan 2026) — superseded (issue #1035)
2026	306	Verified in-force snapshot (corrected, validated against vbpl.vn) — Constitution + 305 Laws/Codes

Features

Feature	Type	Description
id	string	Document number (Số hiệu, e.g. 91/2015/QH13) for the 2026 split — the Constitution uses Hiến pháp 2013; slug for 2026_01/2023/2021
filename	string	Source filename (slug)
title	string	Vietnamese title
type	string	code, law, or constitution
content	string	Full text (cleaned Markdown)
content_length	int32	Character count

The 2026 split — corrected (issue #1035)

An earlier build of the 2026 corpus contained duplicated and mismatched entries (issue #1035). The current 2026 split is a clean rebuild; the original January 2026 release is preserved unchanged as 2026_01.

	2026_01 (original)	2026 (corrected)
Entries	318	306
Duplicate Số hiệu	80	0
Fabricated / non-existent	37	0
Constitution / Codes / Laws	—	1 / 6 / 299

Additional files

• metadata/2026_sources.json — per-document provenance: direct links to 6 sources (≈8 links/doc).
• database/vn_law.sqlite — reference catalogue (laws, source links, in-force status, reconciliation).
• research/ — a comparison with th1nhng0/vietnamese-legal-documents.

Dataset Creation

Curation Rationale

The goal is a clean, verified, currently-in-force corpus of Vietnamese Laws and Codes suitable for fine-tuning and evaluation, distinct from broad raw snapshots. The 2026 split was rebuilt to remove the duplicated/mismatched entries reported in issue #1035.

Source Data

Data Collection and Processing

Full text is sourced from the official vbpl.vn gateway and cross-referenced against five further sources — congbao.chinhphu.vn (Official Gazette), vanban.chinhphu.vn (Government), vietlaw.quochoi.vn (National Assembly), and the reference aggregators thuvienphapluat.vn and luatvietnam.vn. Processing: HTML → Markdown cleaning, de-duplication to one entry per Số hiệu, an in-force filter (status Còn hiệu lực ∪ Hết hiệu lực một phần), and validation that each body matches its requested law.

Who are the source data producers?

The documents are produced by the National Assembly of Vietnam (Quốc hội) and published by the Ministry of Justice (vbpl.vn) and the Government Office (Công báo).

Annotations

No manual annotations. The type and in-force status are derived from the official vbpl metadata.

Personal and Sensitive Information

The corpus consists of public legal texts. The only personal names present are those of public officials (e.g. signatories such as the Chairman of the National Assembly) acting in an official capacity. No private personal data is included.

Bias, Risks, and Limitations

• Point-in-time snapshot. The 2026 split reflects in-force status as of 2026-06; legal status changes over time.
• Consolidation. Amendment laws (Luật sửa đổi, bổ sung) are included as separate documents; the base laws are not auto-consolidated with their amendments.
• Residual coverage. A small number of in-force laws could not be re-sourced (very recent laws awaiting publication, or pre-1992 laws with non-standard Số hiệu); these are documented in the reference DB rather than fabricated.
• Scope. Level 2 of the hierarchy only — no sub-law instruments (Decrees, Circulars, …).

Recommendations

Use the 2026 split for current/in-force law; use 2026_01 only for historical reproducibility of the original release. For legally binding use, confirm the current status of any document against vbpl.vn.

Citation

BibTeX:

@dataset{uts_vlc,
  title        = {Vietnamese Legal Corpus (UTS_VLC)},
  author       = {Underthesea NLP},
  year         = {2026},
  publisher    = {Hugging Face},
  url          = {https://huggingface.co/datasets/undertheseanlp/UTS_VLC}
}

Dataset Card Authors

Underthesea NLP — Vu Anh.

Dataset Card Contact

undertheseanlp@gmail.com