How go avoid training with adding not so important intents

When ading an intent like greet I feel I have to train all my stories aditonal with greet and this increases of course the complexity everytime just ading a further intent. But for the greet intent which is not so important it distroys my stories when uttering in between a greet intent.

How to handle such cases without blowing up your training data?

Hey Datisto!

For simple intents, you can just write a separate 2-line story like

  • greet
    • utter_nice_to_meet_you

and the in-built data-augmentation will glue this conversation onto other stories during training, so the bot will learn to predict the same utterance no matter the context. Another solution for you may be in writing checkpoints. There’s some good information on the docs page for stories so I’d advise you to check it all out :blush:

Hi Tom is there any documentation preserved that references how to create more complex stories in Rasa? I’m at a loss as all documentation links I’m finding to more complex flows has been wiped out.