Reinforcement Learning of Multi-Domain Dialog Policies Via Action Embeddings.
Jorge A. MendezAlborz GeramifardMohammad GhavamzadehBing LiuPublished in: CoRR (2022)
Keyphrases
- multi domain
- reinforcement learning
- optimal policy
- role based access control
- fitted q iteration
- action selection
- markov decision process
- policy search
- cross domain
- action space
- spoken dialogue systems
- state space
- reward shaping
- domain specific
- search computing
- markov decision processes
- function approximation
- reward function
- heterogeneous networks
- markov decision problems
- reinforcement learning algorithms
- partially observable markov decision processes
- control policy
- dimensionality reduction
- active learning
- natural language
- machine learning
- user interface
- transfer learning
- business models
- learning process
- supervised learning
- agent learns
- information extraction
- dynamic programming