How is the next step in a story identified?

Rasa Open Source
1

Hi, I’m trying to understand how the next step is identified. I have two story paths for which some steps are repetitive like:

story path 1: 
intent: intent_1
action: form_1
active_loop: form_1
active_loop: null
action: action_1

story path 2:
intent: intent_2
action: form_1
active_loop: form_1
active_loop: null
action: action_2

What I observed is that the bot identified intent_1, activated form_1 and in the end, triggered action_2. I’m trying to understand if predictions are made at every step of the story?

P.S:

Rasa Version      :         2.8.13
Minimum Compatible Version: 2.8.9
Rasa SDK Version  :         2.8.4
Rasa X Version    :         None 
Python Version    :         3.8.8
Operating System  :         Windows-10-10.0.19041-SP0
2

@siriusraja any thoughts?

3

Hi @lis

Yes, at each turn, the next action is predicted based on the previous conversation history along with how you have defined your policies in your configuration.

Request you to go through the below link to understand the different policies and how it works.

Dialogue Policies

1 Like
4

Hi , my understanding is first intent will be identified with classifier used in “config.yml”. (here diet classifier ).Which you can say indent detection/prediction after that a repetitive loop will be activated as a form

image
image551×688 32.2 KB

5

Hi @siriusraja, thank you! I have set the max_history property for both Memoization Policy and TED Policy. I assume this means it can remember the previous <max_history> things that happened before. In this case, why would it move to action_2 and not action_1? Any idea?

6

Hi @lis

It would depend on your max history value and number of story examples.

1 Like