judicialmind/india-acts

India Acts — Central & State Statutes

A comprehensive corpus of Indian legislation in PDF form — covering both Central (Parliament) Acts and State / Union Territory Acts — in English and Hindi, scraped and consolidated from publicly available government sources (primarily the India Code portal and individual State legislature websites).

This dataset is intended as a research and AI-training resource for tasks such as legal document retrieval, statutory question-answering, summarization, OCR/parsing benchmarks, multilingual legal NLP, and citation analysis.

• Total files: 12,102 PDFs
• Total size: ~21.7 GB
• Languages: English (en), Hindi (hi)
• Coverage: All 28 States + 8 Union Territories of India
• Year range (Central): 1836 – 2025
• Maintainer: JudicialMind

---

📁 Folder Structure

india-acts/
├── central/
│   ├── english/<year>/<act_title>/<file>.pdf
│   └── hindi/<year>/<act_title>/<file>.pdf
└── state/
    ├── english/<State_or_UT>/<year>/<act_title>/<file>.pdf
    ├── hindi/<State_or_UT>/<year>/<act_title>/<file>.pdf
    └── _metadata/scrape_info.json

• Each act lives in its own folder (<sanitized_act_title>/) and contains one or more PDFs (the act itself, amendments, OCRed versions, translations, etc.).
• <year> is the year of enactment (or the year the statute was assigned by the source portal).
• File names typically follow source conventions (e.g. A2024-01.pdf for Act #1 of 2024 on India Code).

---

📊 Dataset Statistics

Central (Parliament) Acts

Language	Files	Size	Year Range
English	891	420 MB	1836 – 2025
Hindi	702	555 MB	1838 – 2024
Total	1,593	975 MB

State & Union-Territory Acts

State / UT	English	Hindi
Andaman and Nicobar Islands	69	0
Andhra Pradesh	265	0
Arunachal Pradesh	152	0
Assam	1,078	0
Bihar	334	11
Chandigarh	396	0
Chhattisgarh	472	47
Dadra & Nagar Haveli and Daman & Diu	228	0
Delhi	137	6
Goa	177	0
Gujarat	250	0
Haryana	275	31
Himachal Pradesh	284	6
Jammu and Kashmir	137	0
Jharkhand	160	2
Karnataka	354	0
Kerala	460	2
Ladakh	63	0
Lakshadweep	3	3
Madhya Pradesh	253	11
Maharashtra	492	38
Manipur	221	3
Meghalaya	105	0
Mizoram	114	0
Nagaland	241	0
Odisha	373	1
Puducherry	107	0
Punjab	391	0
Rajasthan	481	145
Sikkim	113	0
Tamil Nadu	356	2
Telangana	290	0
Tripura	99	0
Uttar Pradesh	265	202
West Bengal	804	0
TOTAL	9,999	510

State English corpus ≈ 19.3 GB, State Hindi corpus ≈ 0.8 GB.

Hindi coverage at state level is uneven — Hindi-belt states (Rajasthan, Uttar Pradesh, Chhattisgarh, Haryana, Maharashtra, Madhya Pradesh, Bihar) provide the bulk of Hindi PDFs; many southern, north-eastern and UT portals publish only English.

---

🚀 Loading the Dataset

Option 1 — Direct file URLs (best for web apps / static sites)

Any file is reachable via:

https://huggingface.co/datasets/judicialmind/india-acts/resolve/main/<path>

Example:

https://huggingface.co/datasets/judicialmind/india-acts/resolve/main/central/english/2024/1_The_Public_Examinations__Prevention_of_Unfair_Mean/A2024-01.pdf

Option 2 — huggingface_hub (Python)

from huggingface_hub import list_repo_files, hf_hub_download

REPO = "judicialmind/india-acts"

files = list_repo_files(REPO, repo_type="dataset")
print(len(files), "files")

local = hf_hub_download(
    repo_id=REPO,
    repo_type="dataset",
    filename="central/english/2024/1_The_Public_Examinations__Prevention_of_Unfair_Mean/A2024-01.pdf",
)
print(local)

Option 3 — Snapshot a subset

from huggingface_hub import snapshot_download
snapshot_download(
    repo_id="judicialmind/india-acts",
    repo_type="dataset",
    allow_patterns=["central/english/**"],
    local_dir="./central_english",
)

Option 4 — Git (full clone, ~22 GB)

git lfs install
git clone https://huggingface.co/datasets/judicialmind/india-acts

Option 5 — List files via REST API

GET https://huggingface.co/api/datasets/judicialmind/india-acts/tree/main?recursive=true

---

🛠️ Data Provenance & Methodology

1. Sources: India Code (central acts and most state portals exposed there) and supplementary scraping of individual state legislature / law-department websites.
2. Scrape window: files were collected through early 2026; an act enacted after the scrape date may be missing.
3. De-duplication: identical files appearing in both a yearly index and the language-specific index were de-duplicated by content hash before publication.
4. Filename normalisation: three PDFs whose original Hindi/Tamil filenames exceeded the 255-byte filesystem limit were renamed to hindi_title.pdf / 1948_tamil_title.pdf while preserving the act folder name.
5. No OCR / no extraction has been applied here — files are the original government PDFs (some are scanned, some are born-digital).

state/_metadata/scrape_info.json retains crawl metadata (URLs, timestamps) for the state acts.

---

⚠️ Known Limitations

• A small number of acts on the source portal exist only as scanned images — text-extraction quality will vary.
• Hindi state coverage is sparse for non-Hindi-speaking states (only the originating state publishes the Hindi text).
• Act folder names are sanitised (special characters replaced with _); the original act title is best read from the PDF's first page.
• This corpus reflects acts as published by source portals at scrape time; amendments enacted after collection are not included.
• This is not legal advice — for authoritative current law always consult the official Gazette of India or the relevant State Gazette.

---

📜 License & Re-use

The PDFs are works of the Government of India and the State Governments. Indian government works are generally available for reuse under Government of India terms (often equivalent to a permissive open-government licence), but you should review the source portal's own terms before commercial redistribution.

This dataset compilation itself is shared without additional restrictions for research and educational use.

---

🧑‍⚖️ Citation

@misc{judicialmind_india_acts_2026,
  title  = {India Acts: Central and State Statutes Corpus},
  author = {JudicialMind},
  year   = {2026},
  url    = {https://huggingface.co/datasets/judicialmind/india-acts}
}

🔗 Related

• 🤗 judicialmind/legal-training-dataset — multilingual legal query/passage training data (3.69 M pairs, 35 languages).