Fallback is triggered even if intent is predicted correctly

I have a fallback route, which should be triggered, if the NLU is not confident enough about a prediction:

  • name: “RulePolicy” core_fallback_threshold: 0.3 core_fallback_action_name: “action_fallback” enable_fallback_prediction: True

Then I have intents like personal_information/ask_name and personal_information/where_from. Whenever I type something to trigger these intents, the fallback route is triggered, even tho it is predicting the intent correctly. I can also see the correct answer selected, but why is it routing to the fallback instead of resonding correctly?

‘intent’: {‘id’: -6425498115225026284, ‘name’: ‘personal_information’, ‘confidence’: 0.9999993443489075}, ‘entities’: , ‘text’: ‘what is your name?’, ‘message_id’: ‘38db3994e5214d7b ab85e5e4a8ccc040’, ‘metadata’: {},

Other intents like greet work as expected.

Hi Valentin,

Seems like you’re using retrieval intents for personal information intents. Wanted to double check you wanted to use that first. If so, can you double check the syntax for your personal_information rule as well as how it’s defined in both your domain and config.yml by following that link above?

If none of that works, can you please share your config.yml and the rule for personal_information?

Thanks!

Oh, it looks like I forgot the rule, and used stories instead. It’s working now, thank you

1 Like