Ideally you should have a full story regarding your entire scenario
I suppose your story looks like an insurance questionnaire , make sure all the steps are covered in the story
* ask_for_insurance
- utter_check_married
* yes (assuming you have a payload here coming from a button)
- utter_check_have_child
* yes
- utter_check_car
* yes
- utter_check_household_owner
* yes
- do_next
Your entire story should be in the training data so it creates an exact blueprint of what an ideal scenario looks like
Agent predicts next action when the confidence for next action between Memoization(confidence either 1 or 0), ML policy(between 0 and 1) and makes the max out of all policies. whichever wins is your next action.
so your ideal scenario should always be there in your training data just that you achieve the next action using Memoization because then the predict is going to be the highest also most accurate.
@souvikg10 What happen when user is saying “yes” before “ask_for_insurance” . How to handle it and return default fallback. I stucked in same scenario .