rmignone/HUDOC-ESC
Dataset Card for the HUDOC European Social Charter (ESC) Corpus Dataset Summary The HUDOC European Social Charter (ESC) Corpus is a comprehensive, bilingual (English and French) dataset containing the decisions, conclusions, and metadata of the European Committee of Social Rights. The data is sourced directly from the Council of Europe's official HUDOC ESC database. This dataset maps the complex, fragmented API responses into a clean, unified schema. Documents are… See the full description on the dataset page: https://huggingface.co/datasets/rmignone/HUDOC-ESC.
Dataset Card for the HUDOC European Social Charter (ESC) Corpus
Dataset Description
- Point of Contact: rahele.mignone@unito.it
Dataset Summary
The HUDOC European Social Charter (ESC) Corpus is a comprehensive, bilingual (English and French) dataset containing the decisions, conclusions, and metadata of the European Committee of Social Rights. The data is sourced directly from the Council of Europe's official HUDOC ESC database.
This dataset maps the complex, fragmented API responses into a clean, unified schema. Documents are grouped by their core identifier, providing English and French translations (titles and full texts) alongside 24 highly detailed legal metadata fields on a single row.
Supported Tasks and Leaderboards
- `text-classification`: Classifying documents by article (
escarticle), state party (escrawstateparty), or document type (escdctype). - `translation`: Serving as a high-quality, domain-specific comparable/parallel corpus for EN-FR legal translation tasks.
- `document-retrieval` / `question-answering`: Building RAG (Retrieval-Augmented Generation) pipelines for human rights law, specific to the European Social Charter.
Languages
The dataset contains texts in English (en) and French (fr).
Dataset Structure
Data Instances
Each row in the Parquet file represents a single, unique base document. If the document exists in both English and French, both versions of the text and title are included in the same row.
An example instance looks like this:
{
"base_id": "XXII-4/def/HRV/8/1",
"api_id_en": "XXII-4/def/HRV/8/1/EN",
"api_id_fr": "XXII-4/def/HRV/8/1/FR",
"title_en": "Conclusions XXII-4 (2021) - Croatia - Article 8-1",
"title_fr": "Conclusions XXII-4 (2021) - Croatie - Article 8-1",
"text_en": "[Full English document text...]",
"text_fr": "[Full French document text...]",
"available_languages": ["EN", "FR"],
"escarticle": "8",
"escrawstateparty": "Croatia",
"escdatedec": "2021-12-07T00:00:00Z",
"escdctype": "Conclusions"
// ... plus 20 other metadata fields
}