incognito-anonymizer 1.4.6

A module to anonymize french text data

Incognito

Description

Incognito is a Python module for anonymizing French text. It uses Regex and other strategies to mask names and personal information provided by the user.
This module was specifically designed for medical reports, ensuring that disease names remain unaltered.

---

Installation

From pip

pip install incognito-anonymizer

From this repository

1. Clone the repository:

   git clone https://github.com/Micropot/incognito
   

2. Install the dependencies (defined in pyproject.toml):

   pip install .
   

---

Usage

Python API

Example: Providing Personal Information Directly in Code

from . import anonymizer

# Initialize the anonymizer
ano = anonymizer.Anonymizer()

# Define personal information
infos = {
    "first_name": "Bob",
    "last_name": "Jungels",
    "birth_name": "",
    "birthdate": "1992-09-22",
    "ipp": "0987654321",
    "postal_code": "01000",
    "adress": ""
}

# Configure the anonymizer
ano.set_info(infos)
ano.add_analyser('pii')
ano.add_analyser('regex')
ano.add_analyser('lossy') # trigger a warning. See doc string for better understanding
ano.set_mask('placeholder')

# Read and anonymize text
text_to_anonymize = ano.open_text_file("/path/to/file.txt")
anonymized_text = ano.anonymize(text_to_anonymize)

print(anonymized_text)

Example: Using JSON File for Personal Information

from . import anonymizer

# Initialize the anonymizer
ano = anonymizer.Anonymizer()

# Load personal information from JSON
infos_json = ano.open_json_file("/path/to/infofile.json")

# Configure the anonymizer
ano.set_info(infos_json)
ano.add_analyser('pii')
ano.add_analyser('regex')
ano.set_mask('placeholder')

# Read and anonymize text
text_to_anonymize = ano.open_text_file("/path/to/file.txt")
anonymized_text = ano.anonymize(text_to_anonymize)

print(anonymized_text)

Example: Annote a file

from . import anonymizer

# Initialize the anonymizer
ano = anonymizer.Anonymizer()

# Load personal information from JSON
infos_json = ano.open_json_file("/path/to/infofile.json")

# Configure the annotator
ano.set_info(infos_json)
ano.add_analyser('pii')
ano.add_analyser('regex')
ano.set_annotator('placeholder')

# Read and annotate text
text_to_anonymize = ano.open_text_file("/path/to/file.txt")
annotated_text = ano.annotate(text_to_anonymize)

print(annotated_text)

Command-Line Interface (CLI)

Basic Usage

python -m incognito --input myinputfile.txt --output myanonymizedfile.txt --strategies mystrategies --mask mymasks

Find Available Strategies, Masks and Annotator

python -m incognito --help

Anonymization with JSON File

python -m incognito --input myinputfile.txt --output myanonymizedfile.txt --strategies mystrategies --mask mymasks json --json myjsonfile.json

To view helper options for the JSON submodule:

python -m incognito json --help

Anonymization with Personal Information in CLI

python -m incognito --input myinputfile.txt --output myanonymizedfile.txt --strategies mystrategies --mask mymasks infos --first_name Bob --last_name Dylan --birthdate 1800-01-01 --ipp 0987654312 --postal_code 75001

To view helper options for the "infos" submodule:

python -m incognito infos --help

Annotation

python -m incognito --input myinputfile.txt --output annotationfile.ann --strategies mystrategies --annotate myannotator infos --first_name Bob --last_name Dylan --birthdate 1800-01-01 --ipp 0987654312 --postal_code 75001

---

Unit Tests

Unit tests are included to ensure the module's functionality. You can modify them based on your needs.

To run the tests:

make test

To check code coverage:

make cov

---

Anonymization Process Details

Regex Strategy

One available anonymization strategy is Regex. It can extract and mask specific information from the input text, such as:

• Email addresses
• Phone numbers
• French NIR (social security number)
• First and last names (if preceded by titles like "Monsieur", "Madame", "Mr", "Mme", "Docteur", "Professeur", etc.)

For more details, see the RegexStrategy class and the self.title_regex variable.

PII Stategy

This strategy is used to catch the personal informations of the patient.

You can use it in CLI with the infos or in a json fil.

For further example you can see the CLI chapter

Lossy Strategy

Another available anonymization strategy is Lossy. The idea is to mask pattern like DUPONT Marc or Marc DUPONT.

!!!warn

It can produce false positive. Be aware that this strategy can can unexpected matched and  loose informations in your text

Get the matched entities

If you want to print the matched entities to check what the code did you can use the get_entities() function

    ano = Anonymizer()
    ano.add_analyzer("regex")
    ano.add_analyzer("lossy")
    ano.set_mask("placeholder")
    ano.anonymize(input)
    entities = ano.get_entities()

The output will match this kind of list : [ {"original": "DUPONT", "replacement": "<NOM>", "type": "NOM", "start": 42, "end": 49}, {"original": "01/01/1970", "replacement": "<DATE>", "type": "DATE", "start": 80, "end": 90}, ]

For more details, see the LossyStrategy class

Anotation Process Details

Standoff Strategy

You can create an annotation file based on the Standoff format.

This file will be automatically created based on the matched entity.

You can find example in the CLI/API chapters

---

License

This project is licensed under the terms of the MIT License.

---

Contributors

• Maintainer: Micropot
  Feel free to open issues or contribute via pull requests!