An enhanced version of a binary scam classification dataset expands to 5 multi-class categories for granular detection. The original dataset contains 14,000 rows of SMS and email-style messages from an Indian context, focusing on banks, UPI, Aadhaar, and government agencies. The dataset was created by Shade63 and last updated on Hugging Face in May 2026.
Use Cases
- Train multi-class text classifiers for scam detection based on the described 5 categories.
- Benchmark fraud detection models on SMS and email-style messages from an Indian context.
- Analyze linguistic patterns of scams targeting Indian banks, UPI, and government agencies.
- Develop domain-specific NLP applications for financial security in India.
Strengths
- The dataset is an enhanced version with 5 multi-class categories, offering more granularity than binary classification.
- The original dataset contains 14,000 rows of SMS and email-style messages.
- The data is domain-specific, focusing on Indian banks, UPI, Aadhaar, and government agencies.
Limitations
- Column-level documentation is absent; field semantics must be inferred after download.
- Row count for the updated multi-class version is unknown, which may limit suitability assessment.
- The dataset may reflect geographic and contextual bias inherent to its Indian-specific source.
Provenance
- Source
- Hugging Face, uploaded by author Shade63.
- Freshness
- Last updated 2026-05-14 16:29:34; freshness should be verified.
- Geography
- India