melikegks/turkish-pii-guard-qwen2.5-1.5b
melikegks/turkish-pii-guard-qwen2.5-1.5b is a 1.5 billion parameter language model developed by melikegks, fine-tuned on Qwen/Qwen2.5-1.5B-Instruct. It is specifically designed for detecting and masking 50 types of personal identifiable information (PII) in Turkish texts, including TCKN, IBAN, phone numbers, names, addresses, and financial details. The model supports both full PII masking and selective masking based on instructions, operating with a context length of 32768 tokens.
Loading preview...
Turkish PII Guard 1.5B Overview
melikegks/turkish-pii-guard-qwen2.5-1.5b is a specialized language model built upon the Qwen/Qwen2.5-1.5B-Instruct base model. Its primary function is to identify and mask Personal Identifiable Information (PII) within Turkish text.
Key Capabilities
- Extensive PII Detection: Supports the detection of 50 distinct types of personal data, including Turkish ID numbers (TCKN), IBANs, phone numbers, full names, addresses, and various financial details.
- Flexible Masking: Capable of masking all detected PII in a given text or performing selective masking based on specific user instructions.
- Instruction-Based Operation: Users can provide instructions to guide the masking process, allowing for nuanced control over which PII types are masked.
Evaluation and Limitations
Evaluated on a Turkish PII masking benchmark, the 1.5B model achieved a schema-neutral exact match score of 0.857. It is noted that the model may exhibit errors in certain selective masking scenarios, particularly when instructed to leave specific PII types unmasked. For high-risk or regulated workflows, independent verification of the model's output is recommended.
Good for
- Automated PII redaction in Turkish documents.
- Developing applications requiring privacy protection for Turkish text data.
- Research and development in Turkish natural language processing focused on data anonymization.