Spam detection in email messages is a critical task in maintaining the effectiveness of email communication. This project leverages deep learning techniques, specifically Word2Vec combined with BiLSTM (Bidirectional Long Short-Term Memory) and XLNet, to develop a system that accurately classifies emails as spam or non-spam (ham). Word2Vec generates word embeddings that capture semantic information from the text, which are then processed using BiLSTM for context understanding in both directions. XLNet, a transformer-based model, enhances this by capturing dependencies and relationships across the entire email. The system is trained on a dataset of 190K labeled emails, enabling it to generalize well for various email messages. This model offers an efficient and reliable approach to email classification, improving spam detection accuracy. The application of these advanced algorithms ensures an intelligent, automated classification of email messages, providing users with a powerful tool for filtering unwanted messages.

Keywords: Spam detection, email classification, Word2Vec, BiLSTM, XLNet, machine learning, deep learning, semantic embeddings, transformer, email filtering.