Proposing Appropriate SMS Spam Detection Approaches for Variations of the Vietnamese Language

This paper introduces suitable methods for detecting SMS spam in various forms of the Vietnamese language. The researchers conducted experiments using five algorithms: SVM, Naive Bayes, Random Forests, CNN, and LSTM, across three different Vietnamese datasets. The results indicate that LSTM and CNN, complemented by a transformer learning model called PhoBert, outperformed traditional machine learning models. Specifically, the LSTM model achieved the highest accuracy of 97.77% when applied to the Vietnamese dataset with full diacritics. Similarly, the CNN model and PhoBert model achieved the highest accuracy of 95.56% when working with the Vietnamese dataset without diacritics.