Emotion Detection Using Transformer Model with Deep Learning

Abstract

Divyang Bharatbhai Joshi, Pankaj Agrawal, Apoorva Ashokbhai Dhokai, Amitkumar Rajpoot, Nandini Pavankumar Agrawal and Avinash Sood

Emotion detection is essential in NLP for classifying emotions from text. Traditional models like SVMs and RNNs struggle with contextual understanding, while Transformer-based models such as BERT and RoBERT offer significant improvements. This study proposes a Transformer-based deep learning approach for emotion classification into six categories. The dataset undergoes preprocessing and label encoding for better model efficiency. Our model achieves over 90% accuracy, surpassing previous deep learning methods (70-85% accuracy). Results show that Transformers capture semantic nuances better than traditional models. This approach is valuable for sentiment analysis, mental health monitoring, and human-computer interaction. Our findings highlight the superiority of Transformer models in emotion detection tasks.

PDF