Audio-Textual Emotion Recognition Using Pre-trained Models: Investigating Various Representations and Fusion Techniques

Key Points

• Emotion recognition improves with the integration of various data representations and fusion techniques.
• Significant advancements were noted, with performance metrics showing a 15% increase over traditional methods.
• Assessment using pre-trained models highlights the importance of combining audio and text data for better accuracy.
• The findings suggest that enhancing emotion recognition technologies could lead to more interactive and empathetic AI systems.