LeMoussel's picture
Upload README.md with huggingface_hub
0fc0aef verified
metadata
dataset_info:
  features:
    - name: source
      dtype: string
    - name: id
      dtype: string
    - name: date_debut
      dtype: string
    - name: date_fin
      dtype: string
    - name: url
      dtype: string
    - name: extra
      dtype: string
    - name: quality_signals
      dtype: string
    - name: text
      dtype: string
  splits:
    - name: train
      num_bytes: 38074982
      num_examples: 7366
  download_size: 8660762
  dataset_size: 38074982
configs:
  - config_name: default
    data_files:
      - split: train
        path: data/train-*
language:
  - fr
task_categories:
  - text-generation
  - text2text-generation
tags:
  - legal
  - text-generation
  - conditional-text-generation
size_categories:
  - 1K<n<10K
license: cc-by-4.0

Légifrance Legislative Text Dataset

Dataset Description

The Légifrance Legislative Text Dataset is a structured collection of French legislative and regulatory texts extracted from the Légifrance platform. This dataset provides machine-readable access to French legal documents, with particular focus on maintaining the integrity of French linguistic features while providing additional metadata and quality signals.

This dataset consists of open data from the Légifrance website, provided by the Tricoteuses project - codes/code_de_la_securite_sociale.

Languages

French (fr)

Intended Uses & Limitations

Intended Uses

  • Legal text analysis and research
  • Natural Language Processing tasks on French legislative documents
  • Legal information retrieval systems
  • Analysis of French regulatory frameworks

Limitations

  • Limited to French legislative texts
  • Dependent on the structure of source Légifrance documents
  • Quality of text extraction depends on the consistency of source markdown formatting

Dataset Structure

Data Fields

  • source: string - Source of the text (e.g., "Code de la sécurité sociale")
  • id: string - Unique identifier of the legislative text
  • date_debut: string - Corresponds to the effective date of the article.
  • date_fin: string - Indicates the date on which the article will be deleted or replaced.
  • url: string - Direct link to the text on Légifrance
  • extra: JSON string containing:
    • État: string - Status of the text
    • Type: string - Type of legislative text
  • quality_signals: JSON string containing:
    • character_count: Total number of characters
    • word_count: Total number of words
  • text: string - The main content of the legislative text

Remarks:

  • For an article currently in force, with no scheduled date for its repeal or replacement, date_fin is set to '2999-01-01'.

  • For an article whose enforcement is deferred without a specific date, date_debut is then assigned '2222-02-22'.

  • If date_fin is specified and differs from '2999-01-01', the reference URL for the upcoming version of the article is constructed as id/date_fin.

Data Splits

The dataset is provided as a single split without train/validation/test partitioning.

Dataset Creation

Source Data

The data comes from the availability of legal texts as open data, retrieved by the tricoteuses-legifrance project.

The dataset is created from Markdown files containing French legislative texts, each with YAML front matter metadata.

Considerations for Using the Data

Social Impact

  • Improves accessibility to French legislative texts
  • Enables computational analysis of legal documents
  • Supports transparency in legal research

Legal Considerations

  • Usage should comply with Légifrance terms of service
  • Attribution should be provided to the original source
  • Users should verify the current validity of legal texts

Additional Information

Dataset Curators

This dataset is programmatically curated from official Légifrance sources.

Licensing Information

Users should refer to Légifrance's licensing terms for the original content.

Citation Information

When using this dataset, please cite both:

  1. The original Légifrance source
  2. This dataset processing implementation

Contributions

Contributions to improve the dataset processing can be made through the repository's issue tracker or pull requests.