Dataset · nvidia
Nvidia synthetic PII/PHI dataset
nvidia/Nemotron-PII
100k English records annotated with 55+ PII/PHI categories for training personal data detection.
What's inside
A person-linked synthetic dataset for training and evaluating PII and PHI detection in text. Contains 100k English records across 50+ industries and span-level annotations for 55+ PII/PHI categories. Generated using NVIDIA NeMo Data Designer on synthetic personas.
What you can do with it
- Train a token-classification model to search for PII/PHI.
- Fine-tune an NER model or LoRA for filtering personal data.
- Evaluate a privacy system on synthetic industry texts.
- Tasks
- Token classification
- Languages
- English
- Size
- 100K–1M rows
- Downloads this month
- 8K
- Likes
- 165
Models trained on it
- OpenMed/privacy-filter-nemotron-v2Token classification 165K
- OpenMed/OpenMed-PII-SuperClinical-Small-44M-v1Token classification 124.8K
- OpenMed/OpenMed-PII-SuperClinical-Large-434M-v1Token classification 118.7K
- OpenMed/privacy-filter-multilingual-v2Token classification 116.2K
- OpenMed/privacy-filter-nemotronToken classification 105.1K
- bardsai/eu-pii-anonimization-multilangToken classification 28.3K
- OpenMed/OpenMed-PII-BioClinicalModern-Large-395M-v1Token classification 18.5K
- OpenMed/OpenMed-PII-NomicMed-Large-395M-v1Token classification 17.7K
How to load it
from datasets import load_dataset
ds = load_dataset("nvidia/Nemotron-PII")