Problem of stories.md

Rasa Open Source
1

I think this is a fundamental question and would like to get some help on it.

Given the following Stories.md from RasaDemo:

## newsletter + affirm feedback
* greet
    - action_greet_user
* signup_newsletter
    - utter_great
    - utter_ask_email
* enter_data{"email": "maxmeier@firma.de"} OR enter_data{"number":"1"}
    - action_store_email
    - slot{"email": "maxmeier@firma.de"}
- action_subscribe_newsletter
    - slot{"subscribed": true}
    - utter_awesome
    - utter_confirmationemail
    - utter_docu
    - utter_ask_feedback
* affirm
    - utter_thumbsup
    - utter_anything_else


## newsletter + deny feedback
* greet
    - action_greet_user
* signup_newsletter
    - utter_great
    - utter_ask_email
* enter_data{"email": "maxmeier@firma.de"} OR enter_data{"number":"1"}
    - action_store_email
    - slot{"email": "maxmeier@firma.de"}
- action_subscribe_newsletter
    - slot{"subscribed": true}
    - utter_awesome
    - utter_confirmationemail
    - utter_docu
    - utter_ask_feedback
* deny
    - utter_thumbsup
    - utter_anything_else

According to Rasa Doc, ## is comment. From my understanding, comment means that this is for developer ONLY, there is NO meaning from the computer point of view.

If it is right, then there are some questions:

  • If a user types something that falls to greet intent, which stories it is fallen into?
  • if user types something that falls to signup_newsletter, and the bot prompts for input, now user types something that falls to deny intent instead of actual input, what would happen?

My question is, since stories.md contains a list of intent and actions, how do Rasa determine which flow it is right now?

If possible, can you please give me a more concrete example on that?

Thank you very much.

2

Your question is about the action prediction. It is key of rasa_core. By default, there are two methods to predict the next action. One is memory prediction, and the other is Keras prediction. For your first confusion, the greet’s action is action_greet_user, whichever it fall to, the response is the same. The story will be stored in a graph, actually, there will only one flow of it. For the another, I think it depends on the Keras model.

Wish to help you.

1 Like