Methodologies for Email Spam Classification using Large Language Models

Email spam classification is an issue that has been around almost since the inception of the email. These tricky emails fool individuals into giving them money, personal information, and more. Several models, such as text classification and Machine Learning, have been used for spam email classification. Despite their beneficial and widespread use, Deep Learning models, which are capable of better understanding the nuances of language, have shown incredible promise in this field; they use a transformer architecture that allows the model to grasp complex language concepts and make relationships and patterns for the data presented. Three methods for email scam classification were demonstrated in this work. Here, we compare class explain ability results from zero-shot summarization, model interpretation via feature importance extraction, and model fine-tuning.