Emotional Language Generation
Natural Language Generation; VAE; GAN; Language Modeling; Reinforcement Learning
Project: Emotion-controllable Chatbot via Conditional Generative Models
The task aims at generating language whose semantics or attributes (e.g., emotional tones) can be controlled. The methods:
- Adversarially trained the sequential data generator with Conditional GAN and VAE
- Using reinforcement learning (RL) to promote the authenticity of the generated language.
- NOTE: This project was done in 2019 and did not use LLMs.
Related Skills
- PyTorch
- Python
- (Controllable) Deep Generative Models (e.g. GAN, VAE)
- Reinforcement Learning
- Language Models
