Deep Reinforcement Learning with Hierarchical Action Exploration for Dialogue Generation.
Itsugun ChoRyota TakahashiYusaku YanaseHiroaki SaitoPublished in: LREC/COLING (2024)
Keyphrases
- action selection
- reinforcement learning
- active exploration
- partially observable domains
- temporal difference
- dialogue system
- action space
- model based reinforcement learning
- exploration strategy
- reward shaping
- function approximation
- multi agent
- generation process
- hierarchical structure
- practical reasoning
- reinforcement learning algorithms
- transition model
- exploration exploitation
- machine learning
- speech acts
- state action
- human machine
- hierarchical model
- model free
- hierarchical models
- dialogue management
- neural network
- state space
- communicative acts
- agent receives
- policy search
- autonomous learning
- multiple layers
- spoken dialogue systems
- human computer
- reward function
- coarse to fine