Dataset · ai4privacy
Multilingual corpus for PII detection and masking
ai4privacy/pii-masking-400k
Hundreds of thousands of texts in six European languages for training models to hide personal information.
What's inside
It contains between 100k and 1 million text examples in English, French, German, Italian, Spanish, and Dutch. The dataset is designed for training and evaluating models that detect and remove personally identifiable information (PII) from texts, tables, and resumes. The authors also note in the description that newer versions with 1.4M and 1.6M examples have already been released.
What you can do with it
- Training token classification models (NER) for text anonymization
- Creating privacy filters before sending data to cloud LLMs
- Tasks
- Text classification, Token classification, Table question answering, Question answering, Zero shot classification, Summarization
- Languages
- English, French, German, Italian, Spanish, Dutch
- Size
- 100K–1M rows
- Downloads this month
- 2K
- Likes
- 64
Models trained on it
- iiiorg/piiranha-v1-detect-personal-informationToken classification 229.2K
- OpenMed/privacy-filter-multilingual-v2Token classification 116.2K
- OpenMed/privacy-filter-multilingualToken classification 70.5K
- vllm-sr/mmbert32k-pii-detector-mergedToken classification 641
- HuggingLil/pii-sensitive-ner-germanToken classification 196
- OpenMed/privacy-filter-multilingual-v2-mlx-8bitToken classification 76
- YATAV-ENT/aegis-personal-pii-nerToken classification 60
- arthrod/gliner-opf-ptbr-pii-v1Token classification 50
How to load it
from datasets import load_dataset
ds = load_dataset("ai4privacy/pii-masking-400k")