Bidirectional Long Term Short Memory Based Text Mining Classification Method for Analysing the What’s App Messages with Text and Emoji

In this digital time, from plain text, emoji’s are crucial to supplement emotional expression. However, effectively understanding and applying text mining techniques to messages containing both text and emoji’s still continues to pose great challenges. This paper proposes a model based on Hybrid CNN-BiLSTM for enhanced emotion recognition for WhatsApp messages. The model utilizes CNN for the extraction of spatial features and BiLSTM for establishing the contextual relation of sequences to text. The framework begins with pre-processing of WhatsApp messages through tokenization of characters, filtering of stop words, normalizing emoji into meaningful representations. The CNN module extracts essential features from input sequences, while the BiLSTM works on learning the context dependency both into the past and into the future. The final classification layer implements a fully connected network with a soft-max activation for sentiment and emotional classification. The model was trained on a diverse and annotated WhatsApp dataset, in which text and emoji patterns are present, rendering it suitable for learning and generalisation tasks. Experimental results have demonstrated that the proposed method performs better than traditional machine-learning techniques in accuracy, precision, recall, and F1-score. The proposed system serves as a quick response to sentiment recognition via WhatsApp, surveillance of social media, and human-computer interaction, grasping the entire complexity of user emotions in digital communication. This study gives further impetus for emoji-aware text mining, providing substantial help within emotion recognition systems under natural language processing.