Constructing narrative using a generative model and continuous action policies.
Emmanouil Theofanis ChourdakisJoshua ReissPublished in: CC-NLG@INLG (2017)
Keyphrases
- generative model
- continuous action
- policy search
- partially observable markov decision processes
- reward function
- probabilistic model
- continuous state
- bayesian framework
- reinforcement learning
- control policies
- em algorithm
- conditional random fields
- prior knowledge
- optimal policy
- dynamic programming
- hidden variables
- semi supervised
- topic models
- markov decision processes
- expectation maximization
- finite state
- decision problems
- reinforcement learning algorithms
- k means
- training data