Phrase-Level Action Reinforcement Learning for Neural Dialog Response Generation.
Takato YamazakiAkiko AizawaPublished in: ACL/IJCNLP (Findings) (2021)
Keyphrases
- reinforcement learning
- fitted q iteration
- action selection
- sensory inputs
- higher level
- levels of abstraction
- markov decision processes
- neural network
- network architecture
- generation process
- learning process
- user interface
- optimal policy
- learning classifier systems
- state action
- learning algorithm
- reward shaping
- partially observable domains
- machine learning