anmolshrivastav/scam-hum-india
Scam/Spam India Dataset A dataset of labeled text messages for scam/spam detection, focused on Indian scam patterns (telecom promotions, lottery fraud, Aadhaar/KYC phishing, UPI/bank fraud, fake job offers, OTP-theft attempts). Dataset Structure Column Type Description text string The message content label string ham (legitimate), spam/scam Total rows: 2272 Label distribution: {'ham': 1377, 'spam': 895} Source Merged from two… See the full description on the dataset page: https://huggingface.co/datasets/anmolshrivastav/scam-hum-india.
Scam/Spam India Dataset
A dataset of labeled text messages for scam/spam detection, focused on Indian scam patterns (telecom promotions, lottery fraud, Aadhaar/KYC phishing, UPI/bank fraud, fake job offers, OTP-theft attempts).
Dataset Structure
- Total rows: 2272
- Label distribution: {'ham': 1377, 'spam': 895}
Source
Merged from two source files (data1.csv, data2.csv) covering real-world Indian SMS/message samples and augmented synthetic examples for underrepresented scam categories (lottery, Aadhaar/KYC, UPI, job scams, OTP phishing). Exact duplicate rows were removed after merging.
Intended Use
Training/fine-tuning binary text classifiers to detect scam or spam messages, particularly ones using India-specific fraud patterns and terminology (Aadhaar, UPI, INR amounts in lakh/crore).
Limitations
- Contains a mix of real and synthetically generated examples; synthetic examples are cleaner/more grammatical than real-world scam text and may not fully capture real-world noise (typos, code-mixing).
- English-language messages only; does not cover Hindi or Hinglish (code-mixed) text.
- Not exhaustive — some scam categories or phrasing patterns may be underrepresented.
Companion Model
This dataset was used to fine-tune `anmolshrivastav/distilbert-scam-detector-india`.