This project is my undergraduate thesis focused on building an empathetic chatbot that generates emotionally intelligent and context-aware responses in Chinese. The system combines the EmpDG (Empathetic Dialogue Generation) model, which focuses on emotion understanding, with a fine-tuned GPT-2 model for natural language generation.
The chatbot is designed to detect the user's emotional state, retrieve relevant commonsense knowledge, and respond with empathy and fluency. By integrating multi-source signals—including conversation history, emotion labels, and external knowledge—the model can produce responses that feel more human and considerate.
Key components include:
- Emotion classification and intent understanding (EmpDG)
- Chinese GPT-2 for fluent and diverse response generation
- Knowledge integration for improved context relevance
- Support for training, evaluation, and interactive chatting
This project demonstrates how empathetic computing and natural language generation can be combined to build emotionally aware dialogue systems. It provides a strong foundation for further research or product development in empathetic AI applications.