Multilingual SMS Spam Classification Using NLP and Transfer Learning

The rapid growth of mobile communication has intensified the spread of SMS spam across multiple languages, posing significant challenges to traditional spam filtering systems that are often language-dependent. This study investigates multilingual SMS spam classification using Natural Language Processing (NLP) techniques combined with transfer learning. By leveraging pre-trained multilingual language models, such as multilingual BERT and related architectures, the proposed approach enables effective knowledge transfer across languages with limited labeled data. The methodology involves text normalization, tokenization, and feature representation using contextual embeddings, followed by fine-tuning on multilingual SMS datasets. Experimental results demonstrate that transfer learning significantly improves classification performance compared to conventional machine learning and monolingual models, particularly for low-resource languages. The findings highlight the scalability, robustness, and practical applicability of multilingual transfer learning frameworks for real-world SMS spam detection systems.