Bangladesh Legal Acts Dataset
Bangladesh's entire statutory framework, 1799 to 2025, structured and annotated with the legal system in force when each act was passed.
What it contains
Why this dataset exists
Legal NLP has rich resources for common-law, English-language jurisdictions and almost none for low-resource, civil-law jurisdictions in South Asia. Bangladesh has more than two centuries of statute law in English, Bengali and both. This dataset makes all of it machine-readable, with the historical context needed to read it properly.
At a glance
| Acts | 1,484 |
| Time span | 1799–2025 (226 years) |
| Sections | 35,633 |
| Footnotes | 14,523 |
| Tokens | about 2.5 million |
| Governments covered | 50 |
| Legal systems covered | 12 |
| Languages | English, Bengali and mixed |
| Licence | CC BY-SA 4.0 |
What each act carries
- Full text, split into titled sections and footnotes
- Act metadata: title, act number, year, publication date, language, token count and repeal status
- Government context: the government in power, its system, the head of government and how they came to power, and their period in office
- Legal-system context: the legal framework, government system, policing system and land relations of the era
- Provenance: the source URL on the official bdlaws.minlaw.gov.bd portal and a fetch timestamp
For example, The Penal Code, 1860 is linked to the British Colonial Government under Lord Canning and to the British Colonial Legal System of 1858–1947.
Files
| File | Contents |
|---|---|
| Contextualized_Bangladesh_Legal_Acts.json | The full corpus with all context attached |
| acts/ | One JSON file per act |
| govt.json | Government lookup table |
| bangladesh_legal_systems.json | Legal-system lookup table |
| filtered_act_list.csv | The act index |
| balanced_1000.csv | A balanced 1,000-act sample |
How it was built
Scraping → cleaning → section-by-section fetching → merging → token counting and language detection → government-context integration → legal-system contextualisation → Bengali-to-English numeral conversion for years → recovery of missing context → final compilation.
Design choices
- Original language preserved. Bilingual and mixed documents are kept as written, because in legal text the wording is the data.
- Built for time, not just search. Linking each act to the legal system in force makes the corpus usable for temporal legal reasoning, not just retrieval.
Cite
BLAD: A Historically Contextualized, Multilingual Dataset of Bangladeshi Legal Acts (1799 to 2025) · arXiv:2607.17111
Dataset DOI: 10.34740/KAGGLE/DSV/12511542 · Code: GitHub
Sections
35,633
Footnotes
14,523
Years
1799–2025
Governments
50
Legal systems
12