Predicted Action is Incorrect

My bot can correctly identify Intent but cannot predict action based on intent sometimes. It predict next action based on policy_1_TEDPolicy but the action is incorrect. What is the best way to resolve it?

Hello @tanzar

You need sufficient training data for TED to learn what to do when. That is, you have to write enough stories.

If you have only a few stories, but many actions, you may also need to increase the number of training episodes.

If you use the Memoization policy instead of TED, it should exactly follow the stories that you provide, but it won’t generalise. Alternatively, if a certain action should always follow from a certain intent (no matter where in the dialogue), then you should define a rule for that.