Hi all,
I have different longer stories while leaving max history for memorization policy rather small as it seemed to make better predictions. When I validate my data with max history I get “of course” a “Story structure conflict after intent” because the max history of 3 or 4 does not see the whole story while there is no story conflict when taking the whole story into account. So now, I was wondering whether I should increase max history of memorization policy (to at least 7) in order to not get the message that there is a story conflict OR should I just leave it to TED to make the right prediction? My goal is to have the best performance as possible.
Here is an example:
- story: happy path intro with chitchat
steps:
- intent: interested
- action: utter_ask_ready
- intent: chitchat
- action: utter_chitchat
- action: utter_ask_ready
- intent: affirm
- action: action_reset_slot
- action: action_ask
- intent: rate
- action: action_set
- action: utter_conditional_response
- action: action_set_internat value
- intent: affirm
- action: action_concrete_action
- intent: affirm
- action: action_concrete_action
- intent: affirm
- action: action_concrete_action
- intent: affirm
- action: action_final
- action: action_return
and
- story: happy path intro with chitchat
steps:
- intent: interested
- action: utter_ask_ready
- intent: affirm
- action: action_reset_slot
- action: action_ask
- intent: rate
- action: action_set
- action: utter_conditional_response
- action: action_set_internat value
- intent: affirm
- action: action_concrete_action
- intent: affirm
- action: action_concrete_action
- intent: chitchat
- action: utter_chitchat
- action: action_concrete_action
- intent: affirm
- action: action_final
- action: action_return
Thank you so much