privacy-kit 0.1.0

PII detection and anonymization using HuggingFace NER models

privacy-kit

PII detection and anonymization for multilingual text, powered by bardsai/eu-pii-anonimization-multilang.

Installation

pip install privacy-kit

Or install directly from GitHub:

pip install git+https://github.com/bards-ai/privacy-kit.git

For development:

git clone https://github.com/bards-ai/privacy-kit.git
cd privacy-kit
uv venv && uv pip install -e .

Quick start

from privacy_kit import PiiModel

model = PiiModel()

text = "Anna Wiśniewska mieszka na ul. Piękna 22, 00-549 Warszawa."

# Simple anonymization
model.anonymize(text)
# → '[PERSON_NAME] mieszka na [LOCATION].'

# Anonymization with entity IDs
model.anonymize(text, mode="ids")
# → {'anonymized_text': '[PERSON_NAME:1] mieszka na [LOCATION:1].',
#    'entities': {'PERSON_NAME:1': 'Anna Wiśniewska',
#                 'LOCATION:1': 'ul. Piękna 22, 00-549 Warszawa'}}

# Extract PII entities
model.extract_pii(text)
# → {'Anna Wiśniewska': 'PERSON_NAME',
#    'ul. Piękna 22, 00-549 Warszawa': 'LOCATION'}

Supported entity types

The underlying model recognizes EU-relevant PII categories including: PERSON_NAME, LOCATION, FINANCIAL_AMOUNT, PERSON_IDENTIFIER, ORGANIZATION_IDENTIFIER, PROPER_NAME, RELIGION_OR_BELIEF, TRADE_UNION_MEMBERSHIP, and more.

License

Apache-2.0