Unable to select response defined in a story

Rasa Open Source
1

I have NLU datasets with size of around 4 MB for FAQ chatbot which has around 159 intents and stories defined for each intents. Most of the stories consists of just one intent and corresponding utterance as response. I have tested the trained model with above NLU and story data using local RASA X environment and I am facing a problem. The problem is an intent is predicted with confidence value of 1 as shown below. troubleshoot1 There is a unique story written for this intent and corresponding utterance as shown below. image

And yet the predicted response is a ‘action_default_fallback’ instead of ‘utter_levfour_empappwork_writecv_response’.

Also, I have defined a rule for nlu_fallback as shown below:

image
image950×146 4.49 KB
and the configuration for fallback threshold is set as shown below.

image

I am wondering why RASA backend is predicting 'action_default_fallback ’ in this case. Could anyone shed some light on this issue, please? I would be grateful for your responses.

Thank you everyone

2

Anything is right for me. Did you put some examples in nlu file for intent levfour_empappwork_writecv_response?

Can you check if has utter_levfour_empappwork_writecv_response in your domain.yml in actions ? Something like that:

actions:
  - utter_levfour_empappwork_writecv_response

Can you make a little test? Increments epochs number (maybe 100) and make new rasa train

This happens using rasa shell too?

My configuration file

pipeline:
  - name: WhitespaceTokenizer
  - name: RegexFeaturizer
  - name: LexicalSyntacticFeaturizer
  - name: CountVectorsFeaturizer
  - name: CountVectorsFeaturizer
    analyzer: char_wb
    min_ngram: 1
    max_ngram: 4
  - name: DIETClassifier
    epochs: 300
    constrain_similarities: true
  - name: EntitySynonymMapper
  - name: ResponseSelector
    epochs: 300
    constrain_similarities: true
  - name: FallbackClassifier
    threshold: 0.7
    ambiguity_threshold: 0.1
policies:
  - name: MemoizationPolicy
  - name: TEDPolicy
    max_history: 5
    epochs: 300
    constrain_similarities: true
  - name: RulePolicy
1 Like
3

Yes, Actually I never though of increasing the epochs for TEDPolicy. I Will definitely give it a try. Thank you for your quick response :slight_smile: