aimtune/az_personal_info_aug_masked

📘 Overview

This dataset consists of augmented Azerbaijani text pairs (clean & masked) that contain personally identifiable information (PII). All content has been automatically generated using ChatGPT to simulate sensitive data scenarios for tasks like PII detection, anonymization, entity masking, and secure data handling.

---

🔍 Dataset Structure

Each example is a paired record:

• original: The full augmented Azerbaijani text containing PII.
• masked: The same text with PII replaced by placeholders (e.g., [NAME], [ADDRESS], [PHONE], etc.).

Example:

{
  "original": "Mənim adım Elçindir və mənim telefon nömrəm +994501234567-dir.",
  "masked": "Mənim adım [NAME] və mənim telefon nömrəm [PHONE]-dir."
}

---

🔐 Masking Rules

The following placeholder tags are used to sanitize the data:

• Full names (first and last): replaced with [NAME]
• Email addresses: [EMAIL]
• Phone numbers (any format): [PHONE]
• Street addresses (including city, state, ZIP/postal code): [ADDRESS]
• Dates of birth: [DOB]
• Social Security numbers, national IDs, passport numbers, or similar identifiers: [ID]
• Credit card numbers and bank account numbers: [PAYMENT_INFO]
• Other directly identifying info (e.g., driver’s license, medical records): relevant placeholder like [MEDICAL_ID]
• Ambiguous cases (e.g., "Dr. Smith"): still masked as [NAME] unless context clearly suggests otherwise

---

🎯 Intended Uses

• Training/evaluating PII detection and masking models
• Research into data anonymity, privacy, and NLP-based sanitization
• Simulating real-world sensitive data scenarios for robust model development

---

🚨 Important Considerations

• ⚠️ Warning: Contains synthetic PII — While the data is auto-generated and not real, it simulates personal data. Treat accordingly.
• Data is not drawn from real individuals, ensuring compliance with privacy concerns.

---

📌 How to Load

from datasets import load_dataset

dataset = load_dataset("aimtune/az_personal_info_aug_masked")
print(dataset["train"][0])

This yields:

{
  "original": "Mənim adım Elçindir ...",
  "masked": "Mənim adım [NAME] ..."
}

---

⚙️ Creation Details

• Entire dataset generated via ChatGPT prompts to produce realistic Azerbaijani sentences with common PII tokens.
• Augmentation includes varied structures, vocabulary, and PII placements to enhance robustness.

---

📑 Citation

If you use this dataset, please cite:

@misc{az_personal_info_aug_masked_pairs,
  title        = {Azerbaijani ChatGPT‑generated PII Augmented Dataset},
  author       = {Hamza Agar},
  howpublished = {Hugging Face Dataset},
  year         = {2025},
  url          = {https://huggingface.co/datasets/aimtune/az_personal_info_aug_masked}
}

---

🔗 Dataset Link

Browse or download the dataset here:

👉 https://huggingface.co/datasets/aimtune/az_personal_info_aug_masked