Multi-intent doesn't work



 language: en 
   - name: SpacyNLP
     model: en_core_web_md
     case_sensitive: False 
   - name: SpacyTokenizer
     intent_tokenization_flag: True
     intent_split_symbol: "+"
   - name: SpacyFeaturizer


 - story: greet goodbye story
   - intent: greet+goodbye
   - action: utter_thankyou

intent files

  • intent: greet examples: |

    • hey
    • hello
    • hi
    • hello there
  • intent: goodbye examples: |

    • good afternoon
    • cu
    • good by
  intent: greet+goodbye
   examples: |
     - Hi. Goodbye
     - Hello and Good Bye
     - Good morning and Good Bye

I have also correct configurations in domain.yml. I have many examples for each intent.

But it doesn’t correctly predict greet+goodbye

What am I doing wrong?

Hi @lmb, it could be a case of imbalance between number of training examples per intent in your NLU data. When you run rasa shell nlu and you type in a test utterance for greet+goodbye, what does the confidence breakdown look like?