Hello, I am using the RASA 1.4.5 version and I am facing this issue. The bot is not predicting action after 3rd step of any story. Instead of giving me the reply of the 4th node, it takes me to fallback action. When I debug it in the console, I found the NLP gives excellent confidence for the 4th node, but core action confidence is met with fallback core confidence. Below are my stories.md file. ## story One * Greeting - greeting_action ## story two * Getting_Started - getting_started_action * Designation - designation_action * bot_type - bot_type_action * connect_with_live_agent - live_agent_action ## story three * Getting_Started - getting_started_action * Designation - designation_action * are_you_bot - are_you_bot_action * affirm - affirm_action ## story four * Getting_Started - getting_started_action * Designation - designation_action * are_you_bot - are_you_bot_action * deny - deny_action * connect_with_live_agent - live_agent_action ## story five * Designation - designation_action * are_you_bot - are_you_bot_action * deny - deny_action * connect_with_live_agent - live_agent_action ## story six * Designation - designation_action * are_you_bot - are_you_bot_action * affirm - affirm_action ## story seven * Designation - designation_action * bot_type - bot_type_action * connect_with_live_agent - live_agent_action ## fallback story * Out_Of_Scope - Fallback_action Below is my config.yml file. language: en pipeline: tensorflow_embedding policies: - name: AugmentedMemoizationPolicy - name: MappingPolicy - name: FallbackPolicy nlu_threshold: 0.4 core_threshold: 0.3 fallback_action_name: "Fallback_action" Below is my domain.yml file. actions: - getting_started_action - designation_action - bot_type_action - live_agent_action - are_you_bot_action - affirm_action - deny_action - Fallback_action - greeting_action intents: - Getting_Started - Designation - bot_type - connect_with_live_agent - are_you_bot - affirm - deny_action - Out_Of_Scope - Greeting slots: slot_Greeting: type: text I want the bot to predict the next node of the story instead of fallback action.

@kulveersingh , you should run rasa interactive to handle this. Doing this, you’ll teach and guide your assistant identify the intents and which actions it should answer. Try this way, hope it helps :smiley:

Bot is not working with long stories

Rasa Open Source

honeydhi (Davinder Singh) January 4, 2021, 7:32am 3

@marcos.allysson Thanks for your response. We have already done the same thing. But did not get any success.

Topic		Replies	Views
RASA ALLWAY PREDICT NEXT ACTION IS FALLBACK ACTION Rasa Open Source	5	806	July 17, 2019
Rasa stops predicting next actions Rasa Open Source	1	319	June 29, 2022
TED policy predicton not consistent Rasa Open Source	1	290	July 12, 2022
Rasa not following story Rasa Open Source	6	3033	December 12, 2018
Rasa X and action_default_fallback Rasa Open Source	6	603	May 18, 2021

Bot is not working with long stories

Related topics