PII Detection Model - Phi3 Mini Fine-Tuned

This model, developed by ab-ai, is a fine-tuned version of the 4 billion parameter Phi3 Mini specifically engineered for detecting Personally Identifiable Information (PII) in text. It leverages the Hugging Face Transformers framework to identify and extract a broad spectrum of PII entities, making it a robust tool for data privacy and compliance applications.

Key Capabilities

• Extensive PII Detection: Capable of recognizing over 50 distinct PII entities across various categories, including:
  • Personal Information: firstname, lastname, dob, age, gender, sex, height, eyecolor.
  • Contact & Address: email, phonenumber, url, street, city, state, zipcode, country.
  • Financial Data: accountnumber, creditcardnumber, iban, bic, currency, amount.
  • Unique Identifiers: ssn, imei, mac, vehiclevin, pin.
  • Cryptocurrency: bitcoinaddress, litecoinaddress, ethereumaddress.
  • Other: ip, password, username, useragent.
• Structured Output: Designed to return detected PII in a JSON format, facilitating easy integration into automated workflows for data processing and redaction.

Use Cases

• Data Redaction: Automatically identify and mask sensitive information in documents and communications.
• Privacy Protection: Enhance data privacy by detecting and managing PII in datasets.
• Compliance: Assist organizations in adhering to data protection regulations by accurately identifying PII.
• Data Anonymization: Prepare datasets for analysis by removing or anonymizing personal identifiers.