In this file we keep track of features we'd like to implement in the Plato Research Dialogue System.
-
Write more parsers for existing datasets
-
Add logging levels
- no logging
- debug
- info
- web ui
-
Add printing levels, same as logging
-
Write scripts for easy experimentation
-
Add end-to-end example system
-
Implement more dialogue policies
-
Have one LudwigModel class instead of one for each core component
-
Support multiple domains
-
Support multiple languages
-
Address all #TODOS in the code
-
At the generic agents, set at the config which agent is in control of the state and therefore responsible for assessing if the agent is at a terminal state or not.