unpii 0.1.0

High-performance French medical text anonymization

unpii

High-performance French medical text anonymization library. Rust core with Python bindings and native Polars integration.

Designed to process millions of documents efficiently.

Installation

pip install unpii

Quick Start

import unpii

text = "Dr Martin au 06 12 34 56 78, email: [e-mail]"

# Mask with placeholders (default)
unpii.mask(text)
# → "Dr <NOM> au <TELEPHONE>, email: <EMAIL>"

# Mask with stars
unpii.mask(text, mask="stars")
# → "Dr ***** au *****, email: *****"

Detection Modes

Two detection levels: standard (reliable patterns) and paranoid (aggressive).

# Standard: titles, known patterns, blacklisted names
unpii.mask("Dr Martin est ici")
# → "Dr <NOM> est ici"

unpii.mask("DUPONT Jean est ici")
# → "DUPONT Jean est ici"  (not detected in standard)

# Paranoid: also catches UPPERCASE Titlecase patterns, 5+ digit sequences, loose emails
unpii.mask("DUPONT Jean est ici", mode="paranoid")
# → "<NOM> est ici"

Ignore Groups

Skip specific categories:

unpii.mask("Dr Martin au 06 12 34 56 78", ignore_groups=["TELEPHONE"])
# → "Dr <NOM> au 06 12 34 56 78"

Inspect Detected Spans

Dry-run mode to see what would be masked:

for span in unpii.find_spans("Dr Martin au 06 12 34 56 78"):
    print(span)
# Span(start=3, end=9, category="NOM")
# Span(start=13, end=27, category="TELEPHONE")

Polars Integration

Native expression plugin — Polars handles parallelization automatically:

import polars as pl
import unpii

df = pl.DataFrame({"text": [
    "Dr Martin au 06 12 34 56 78",
    "Email: [e-mail]",
    "Maladie de Parkinson",
]})

df.with_columns(pl.col("text").unpii.mask())
# ┌─────────────────────────┐
# │ text                    │
# ╞═════════════════════════╡
# │ Dr <NOM> au <TELEPHONE> │
# │ Email: <EMAIL>          │
# │ Maladie de Parkinson    │  ← protected by whitelist
# └─────────────────────────┘

# With options
df.with_columns(
    pl.col("text").unpii.mask(mask="stars", mode="paranoid", ignore_groups=["TELEPHONE"])
)

Categories

Group	Placeholder	Standard	Paranoid
NOM	<NOM>	Titles + name, blacklist	UPPERCASE/Titlecase patterns, initials
TELEPHONE	<TELEPHONE>	French phone numbers	—
EMAIL	<EMAIL>	Valid emails	Anything with @
DATE	<DATE>	DD/MM/YYYY, literal months, ISO	—
BIRTHDATE	<BIRTHDATE>	né(e) le + date	—
ADRESSE	<ADRESSE>	Street number + type + name	—
CODE_POSTAL	<CODE_POSTAL>	5 digits + city name	—
NIR	<NIR>	French social security number	—
IBAN	<IBAN>	French IBAN	—
NUMBER	<NUMBER>	—	5+ consecutive digits

Whitelist

Medical eponyms (Parkinson, Alzheimer, Verneuil...) are protected from masking by a global whitelist, regardless of which group detected them.

API Reference

def mask(
    text: str,
    *,
    mask: str = "placeholder",       # "placeholder" or "stars"
    mode: str = "standard",          # "standard" or "paranoid"
    ignore_groups: list[str] | None = None,
) -> str: ...

def find_spans(
    text: str,
    *,
    mode: str = "standard",
    ignore_groups: list[str] | None = None,
) -> list[Span]: ...

# Span attributes: .start, .end, .category

Performance

Rust core with compiled regex and Aho-Corasick automata. All rules and dictionaries are embedded in the binary — zero I/O at runtime.

Single-threaded: ~10ms per 7KB document (~100 docs/sec). With Polars: automatic parallelization across all cores.

License

MIT