Hi, this is more of a general question about how to write stories:
Should one write the stories always from beginning to end or can you also only model parts of the conversation as a story? The following example might not make a lot of sense but should make the question more clear:
## full story
* greet
- utter_greet
* make reservation
- reservation_form
- form{"name": "reservation_form"}
- form{"name": null}
* ask_question
- utter_answer_question
* bye
- utter_bye
## other combinations of the subconversations
...
OR
## greet
* greet
- utter_greet
## reservation
* make reservation
- reservation_form
- form{"name": "reservation_form"}
- form{"name": null}
## ask_question
* ask_question
- utter_answer_question
## bye
* bye
- utter_bye
I can imagine that the answer to this question depends on the policy used and that the split example is worse in terms of Memoization policy (unless you use Augmentation). On the other hand it reduces the amount of stories you need to model the same conversations a lot, since you dont have to define all combinations explicitly.
Anyway how does splitting the stories affect the performance of the TED policy?