Transforming Bengali Emails: Efficient Multi-Class Classification Using Transformers

Email is vital for effective and professional interaction in both personal and company settings worldwide. As email usage continues to grow, it has become the primary mode of communication for many. Communicating successfully is more important than ever. It requires proper classification, categorization and organization of messages. Classifying emails is an essential research area not only due to the high volume of emails sent daily, but also for saving time and increasing productivity. Despite Bengali being one of the world’s most widely spoken languages, there is a substantial gap in terms of reliable email classification systems. This gap highlights the fundamental requirement for a reliable and effective Bengali email classification system. Our research aims to address this issue by proposing a system to classify Bengali emails into different classes. The outcomes of our study demonstrate that the proposed method can competently classify Bengali emails, allowing users to manage their communications more systematically and efficiently. We have fine-tuned pre-trained transformers on a Bengali email dataset to perform our task, achieving 91% accuracy. This highlights the system’s scalability and effectiveness for real-world applications.