Upload README.md with huggingface_hub
Browse files
README.md
ADDED
|
@@ -0,0 +1,70 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
language:
|
| 3 |
+
- fr
|
| 4 |
+
license: cc-by-nc-4.0
|
| 5 |
+
size_categories:
|
| 6 |
+
- n<1K
|
| 7 |
+
task_categories:
|
| 8 |
+
- text-retrieval
|
| 9 |
+
- question-answering
|
| 10 |
+
tags:
|
| 11 |
+
- legal
|
| 12 |
+
- french-law
|
| 13 |
+
- rag
|
| 14 |
+
- spadalab
|
| 15 |
+
pretty_name: "Certifications Securite"
|
| 16 |
+
---
|
| 17 |
+
|
| 18 |
+
# Certifications Securite
|
| 19 |
+
|
| 20 |
+
Export de la collection certifications_securite
|
| 21 |
+
|
| 22 |
+
**479 chunks** issus de **265 articles uniques** — droit francais, prets pour ingestion RAG.
|
| 23 |
+
|
| 24 |
+
## Format
|
| 25 |
+
|
| 26 |
+
- `data/chunks.parquet` — format principal (Parquet/snappy)
|
| 27 |
+
- `chunks.json` — meme contenu en JSON
|
| 28 |
+
- `datasheet.json` — fiche descriptive complete (Datasheet for Datasets, Gebru et al.)
|
| 29 |
+
- `metadata.json` — statistiques + hash SHA-256 d'integrite
|
| 30 |
+
|
| 31 |
+
Chaque chunk contient au minimum : `article_id`, `titre_article`, `code_source`, `text`, `hierarchie`, `context_prefix`, `nb_tokens_estime`.
|
| 32 |
+
|
| 33 |
+
## Methode d'extraction
|
| 34 |
+
|
| 35 |
+
Extraction automatisee depuis les API officielles (Legifrance PISTE, EUR-Lex) et telechargement de documents PDF publics (ANSSI, ENISA, CNIL).
|
| 36 |
+
|
| 37 |
+
## Chunking
|
| 38 |
+
|
| 39 |
+
Decoupe intelligente par structure juridique native : alineas pour Legifrance, paragraphes numerotes pour EUR-Lex, sous-sections pour les PDFs. Taille max ~1500 tokens, min ~50 tokens.
|
| 40 |
+
|
| 41 |
+
## Licences sources
|
| 42 |
+
|
| 43 |
+
- Licence Ouverte Etalab 2.0 (donnees Legifrance)
|
| 44 |
+
- EUR-Lex — reutilisation libre (Decision 2011/833/UE)
|
| 45 |
+
- Documents PDF publics — reproduction autorisee sauf mention contraire
|
| 46 |
+
|
| 47 |
+
## Licence dataset
|
| 48 |
+
|
| 49 |
+
**CC-BY-NC-4.0** — usage non commercial uniquement. Pour usage commercial : licence custom SpadaLab (en cours).
|
| 50 |
+
|
| 51 |
+
## Charger le dataset
|
| 52 |
+
|
| 53 |
+
```python
|
| 54 |
+
from datasets import load_dataset
|
| 55 |
+
ds = load_dataset("SpadaLab/certifications-securite", data_files="data/chunks.parquet")
|
| 56 |
+
print(ds["train"][0])
|
| 57 |
+
```
|
| 58 |
+
|
| 59 |
+
## Sources principales
|
| 60 |
+
|
| 61 |
+
- Liste de vérification IA — RGPD (CNIL)
|
| 62 |
+
- NIS2 Technical Implementation Guidance v1.0 (ENISA)
|
| 63 |
+
- SecNumCloud v3.2 — Référentiel d'exigences (ANSSI)
|
| 64 |
+
- Cloud Security Guide for SMEs (ENISA)
|
| 65 |
+
|
| 66 |
+
## Contact
|
| 67 |
+
|
| 68 |
+
- Site : https://spadalab.fr
|
| 69 |
+
- Email : contact@spadalab.fr
|
| 70 |
+
- Org HuggingFace : https://huggingface.co/SpadaLab
|