A Federated Learning Approach for Text Classification Using NLP

Abstract

Text classification is important in many aspects of natural language processing (NLP), such as word semantic categorization, emotion analysis, question answering, and conversation management. Law, health, and marketing are just a few of the professions that have made use of it throughout the last century. We focused on emotion analysis in this research, which includes categories like happiness, sadness, and more. We investigated the precision of three alternative models for assessing the emotional tone of written text. Deep learning models like GRU (Gated Recurrent Unit) and CNN (Convolutional Neural Network) are used in conjunction with the Bi-LSTM (Bidirectional LSTM) model. These three major deep learning architectures have been extensively studied for classification applications. Finally, the model with the greatest accuracy on the dataset was trained using federated learning (FL). Using the FL approach, more data has to be collected, eliminating data gaps and ensuring data security. The focus of this research is to increase the model’s accuracy by collaborating with FL approaches while maintaining data confidentiality.