This research introduces a novel Emotion-Aware Embedding Fusion framework that significantly improves the emotional intelligence of large language models for mental health applications.

• Combines multiple emotion lexicons with hierarchical fusion techniques to prioritize emotional features in therapy contexts
• Tests the approach across multiple LLMs including Flan-T5, LLAMA 2, DeepSeek-R1, and ChatGPT 4
• Enables more empathetic and contextually relevant responses critical for automated psychotherapy
• Creates a foundation for more effective mental health chatbots that better understand patient emotions

This breakthrough matters for healthcare because it addresses the critical gap between technical LLM capabilities and the emotional understanding necessary for effective psychiatric support applications.

Emotion-Aware Embedding Fusion in LLMs (Flan-T5, LLAMA 2, DeepSeek-R1, and ChatGPT 4) for Intelligent Response Generation