Datasetpublic

ArthurYeghinyan/arlis-armenia-legal-dataset

Complete Armenian Legislation & Legal Corpus (ARLIS 1990–2026) This repository contains the complete, unified, and cleaned digital corpus of all official legislative, normative, and regulatory acts of the Republic of Armenia published in the Armenian Legal Information System (ARLIS — arlis.am) from 1990 to September 2026. 📊 Dataset Summary Total Legal Acts: 188,894 acts Historical base (1990 – April 2023): 154,934 acts (data/train-historical-part01.parquet –… See the full description on the dataset page: https://huggingface.co/datasets/ArthurYeghinyan/arlis-armenia-legal-dataset.

sourceHugging Faceotherupdated 10d agoView on Hugging Face

0likes224downloads

Dataset Card

Complete Armenian Legislation & Legal Corpus (ARLIS 1990–2026)

This repository contains the complete, unified, and cleaned digital corpus of all official legislative, normative, and regulatory acts of the Republic of Armenia published in the Armenian Legal Information System (ARLIS — arlis.am) from 1990 to September 2026.

📊 Dataset Summary

—Total Legal Acts: 188,894 acts
—Historical base (1990 – April 2023): 154,934 acts (data/train-historical-part01.parquet – part07.parquet)
—Modern updates (April 2023 – September 2026): 33,960 acts (data/train-part1.parquet – part3.parquet)
—Total Parquet Size: ~1.79 GB (compressed with ZSTD)
—Language: Armenian (AM)
—Primary Scope: All Constitutions, Codes (Քրեական, Քաղաքացիական, Աշխատանքային, etc.), Laws, Presidential Decrees, Government Decisions, and Departmental Regulations.

🛠 Harmonized 21-Column Schema

Both historical and modern records have been unified under the exact same schema. Raw HTML formatting has been cleaned into normalized plain text, and laws have been split into granular articles:

Field	Type	Description
`uniqid`	string	Unique document ID in ARLIS (e.g. `6611`, `177003`)
`pdf_link`	string	Direct link to official PDF document
`ActNumber`	string	Official legal act identifier (e.g. `ՀՕ-455-Ն`, `N 218`)
`DocType`	string	Document type (e.g. `Հիմնական`, `Ինկորպորացիա`)
`ActType`	string	Act category (e.g. `Օրենք`, `Որոշում`, `Հրամանագիր`)
`ActStatus`	string	Current legal status (`Գործում է` / `Ուժը կորցրել է`)
`Source`	string	Official gazette publication source
`EnactmentOrgan`	string	Body that enacted the law/decision
`EnactmentDate`	string	Enactment date (`DD.MM.YYYY`)
`EnactmentLocation`	string	City/location of enactment
`SigningOrgan`	string	Signing official / body
`SigningDate`	string	Date of signing (`DD.MM.YYYY`)
`RatificationOrgan`	string	Ratifying organ
`RatificationDate`	string	Ratification date (`DD.MM.YYYY`)
`EffectiveDate`	string	Legal effective date (`DD.MM.YYYY`)
`InterruptDate`	string	Date act was repealed/voided (or `None` if active)
`title`	string	Full official title of the legal act
`body`	string	Clean, normalized plain Armenian text (all HTML stripped)
`articles_count`	int64	Count of individual articles found in the document
`language`	string	Language code (`AM`)
`articles_json`	string	Structured JSON array of articles: `[{"article_number", "header", "content"}]`

🚀 How to Load & Use

Using Hugging Face Datasets (Python)

python

from datasets import load_dataset

# Loads all 188,894 acts automatically across all 10 parquet partitions
dataset = load_dataset("ArthurYeghinyan/arlis-armenia-legal-dataset")
print(dataset)

Using Pandas or DuckDB

python

import duckdb

# Query active laws directly over HTTP
con = duckdb.connect()
df = con.execute("""
    SELECT ActNumber, title, EnactmentDate 
    FROM 'hf://datasets/ArthurYeghinyan/arlis-armenia-legal-dataset/data/*.parquet'
    WHERE ActStatus = 'Գործում է' AND ActType = 'Օրենք'
    LIMIT 10
""").df()
print(df)