Abstract

Emotion recognition in social media is a challenging task due to noisy, informal, and highly contextual text. This study presents an improved BERT-based framework for classifying six primary emotions—Happy, Sad, Angry, Fear, Surprise, and Neutral. The methodology has been enhanced through rigorous preprocessing, contextual tokenization, class-imbalance handling using random oversampling, and optimized fine-tuning of BERT. A comprehensive experimental setup was employed, including detailed evaluation metrics, confusion matrix analysis, and performance comparison across varying training configurations. High-resolution figures and expanded result interpretations provide deeper insight into model behavior, particularly for minority classes. The proposed approach demonstrates strong performance on social-media datasets and establishes a foundation for future multimodal fusion techniques involving text, emojis, and visual cues.

References

• J. Devlin et al., “BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding,” NAACL-HLT, 2019.
• A. Vaswani et al., “Attention is All You Need,” NeurIPS, 2017.
• Q. Li et al., “Emotion recognition in social media based on multi-task BERT fine-tuning,” IEEE Access, 2021.
• Q. Li et al., “Fine-tuning BERT for multi-class emotion classification in tweets,” Information Processing & Management, 2022.
• Y. Sun et al., “EmoFusion: A multimodal framework for emotion classification in social media,” Information Fusion, 2023.
• A. Chakraborty et al., “Handling data imbalance in emotion classification using resampling techniques,” IEEE Trans. Affective Computing, 2021.
• S. Alharthi et al., “Sentiment analysis and emotion classification on social media: A survey,” Journal of King Saud University, 2021.
• D. Ghosal et al., “EmotionX: Multimodal emotion recognition using ensemble deep models,” Information Fusion, 2021.
• Y. Zhang et al., “Context-aware emotion recognition from tweets using transformers,” Computational Intelligence, 2022.
• R. Chatterjee et al., “Multimodal sentiment and emotion analysis in social media: A survey,” ACM TOMM, 2023.
• M. Singh et al., “Real-time emotion detection using deep learning,” Neural Computing and Applications, 2023.
• L. Wang et al., “BERT-based ensemble learning for emotion recognition,” IEEE Access, 2023.
• L. Wang et al., “Context-aware multimodal emotion recognition,” IEEE Trans. Affective Computing, 2024.
• Z. Yang and E. Hovy, “Fine-tuning large models for emotion detection,” Journal of AI Research, 2022.
• Y. Zhao et al., “Sentiment and emotion detection from short texts: A survey,” Information Processing & Management, 2022.