RASA Core not giving response despite of having around 30k dataset

Hello all, I have create a custom file to change None to different intent because earlier I was using spacy_sklearn but I needed to integrate dual language bot, i.e english and german. So here is the updated config file :

- name: "WhitespaceTokenizer"
- name: "RegexFeaturizer"
- name: "CRFEntityExtractor"
- name: "EntitySynonymMapper"
- name: "CountVectorsFeaturizer"
- name: "EmbeddingIntentClassifier"
- name: "customNLU.CustomComponent"

After the training, Rasa NLU shows the intent and confidence score perfectly. But by training it for Rasa Core, it doesn’t generate response for any single intent. The custom configuration file was based on :https://github.com/RasaHQ/rasa/issues/2547. Please help.

Hi @LKiera. Can you confirm that your NLU model actually classifies the intents correctly when you test it with rasa shell? Also, I would need more information about the training stories you have - how many of the do you have?

No, unfortunately due to the structure I choose to go ahead with the rasa shell is not able to excute I guess. The training stories is 30000 in length all under single hood. It was more like a loop executed via random function, concatenating the intent with action repeatedly

I suspect the issue could be due to the way the stories are created - they have to reflect the dialogues you would like your assistant to be able to handle and learn from. If the stories are randomly concatenated from the intents and actions, I guess they don’t necessarily reflect the conversations that make sense which makes the model quite confused. I would like to see what happens once you run your bot in a debug mode:

rasa shell --debug

Speaking of Rasa shell, why are you not able to execute it? Are you getting any errors when running it? rasa shell --debug

Seems like there was something needed more then rasa nlu and rasa core. Now the rasa shell is working and yes all the intents shown are proper.