Emotion analysis techniques and development directions in social networks
Jiahui Chen · IET conference proceedings. · 2025
In recent years, deep learning and neural network methods have established themselves as the foremost techniques in emotion analysis. Their powerful abilities in feature extraction and pattern recognition have driven this emergence. Meanwhile, the surge in user-generated content on platforms such as Twitter, Facebook, and Weibo has underscored the need for efficient automatic extraction and analysis of emotional information within this content. This study examines the primary methods used in emotion analysis, focusing on Deep Neural Networks (DNN), Convolutional Neural Networks (CNN), Recurrent Neural Networks (RNN), and multimodal emotion analysis techniques. The research finds that CNNs perform excellently in short text emotion analysis, with an accuracy of 0.91, making them suitable for handling local features. RNNs have advantages in processing long texts, and when combined with other models, they can better address the vanishing gradient problem, achieving an accuracy of 0.90. The integration of multimodal data significantly improves the accuracy of emotion detection. Despite significant technological advances, emotion analysis still faces challenges such as data diversity, implicit emotional expressions, and privacy protection. Future research should further explore multimodal integration and cross-cultural adaptation to enhance the comprehensive performance and application value of emotion analysis.