Ignore intents in Rasa form makes bot behave strange

Rasa Open Source
1

In have defined a Rasa form authenticate_form which excludes 5 different intents by using ignored_intents.

forms:
  authenticate_form:
    ignored_intents:
       - stop_asking_me
       - stop_talking_with_bot
       - where_to_find_contract_no
       - where_to_find_account_id
       - can_i_get_loan
    required_slots:
      name:
        - type: from_text
      name_last:
        - type: from_text
      street_name:
        - type: from_text
      house_number:
        - type: from_text
      city_name:
        - type: from_text
      contract_no:
        - type: from_text
      account_id:
        - type: from_text

Without these 5 intents the form works fine. The input of the user will be mapped as planned. However this changes when I try to exclude specific intents.

My problem is that my Bot behaves very strange.

Example A: If the bot asks for the city_name and the user types Frankfurt-Main, the bot will trigger the intent stop_asking_me. Therefore the slot is not filled with the (correct) slot value.

Example B: If the bot asks for name and the user types “Franci”, the bot will trigger stop_asking_me.

Well, this is obviously wrong, because the name Franci has absolutly nothing to do with the intent stop_asking_me in which the user asks the bot to stop asking further questions.

Personally, I have the feeling the bot doesn’t find the correct intent. However, I have provided about 20-30 examples for each of the 5 intents. Even if the user types almost the exact sentence listed in the training data, the intent is sometimes not recognized.

What can I do so that the bot always triggers the correct intent?

2

Nobody has an idea?

Can it be the model pipeline? At the moment I am using this:

pipeline:
   - name: WhitespaceTokenizer
   - name: RegexFeaturizer
   - name: LexicalSyntacticFeaturizer
   - name: CountVectorsFeaturizer
   - name: CountVectorsFeaturizer
     analyzer: char_wb
     min_ngram: 1
     max_ngram: 4
   - name: DIETClassifier
     epochs: 100
     constrain_similarities: true
   - name: EntitySynonymMapper
   - name: ResponseSelector
     epochs: 100
     constrain_similarities: true
   - name: FallbackClassifier
     threshold: 0.3
     ambiguity_threshold: 0.1

Maybe the training data (20-30 examples per intent) is simply not enough for this pipeline?

3

@Chatbot_Ra Please see this syntax how to add particular intent based on required slots: 

forms:
  your_form:
    required_slots:
        slot_name:
        - type: from_text
          intent: intent_name
          not_intent: excluded_intent

I would encourage you to please see the document of form on rasa or click the above link whilst using ignored_intents, required_slots , intent, rules and stories. Hope you able to solve your issue. Good Luck mate!