BANGALABARTA A BENCHMARK DATASET FOR SPAM AND SMISHING SMS DETECTION IN MOBILE COMMUNICATION

The BangalaBarta dataset presents a comprehensive and linguistically diverse collection of 2,772 Bangla (Bengali-Bangladeshi) SMS messages designed for the detection and classification of spam and smishing (SMS-based phishing) communications. The dataset is categorized into three distinct classes—Smishing, Promotional, and Normal SMS—capturing the heterogeneity of text messages commonly exchanged across major Bangladeshi telecommunication networks including Grameenphone, Banglalink, and Robi. Each message in the dataset is meticulously annotated to reflect its communicative intent, providing a robust foundation for developing and evaluating machine learning and natural language processing (NLP) models in the domain of mobile security and spam filtering. The dataset is distributed in a structured format (CSV), facilitating reproducibility and ease of use in experimental research. BangalaBarta serves as a valuable resource for advancing research in Bangla language processing, spam and smishing detection, and telecommunication security. By offering a representative sample of real-world SMS content, this dataset aims to enhance automated systems capable of safeguarding users against fraudulent and unwanted messages in Bangla digital communication environments.