I have already read every imaginable post about this topic and I can’t know why it doesn’t work.
I thought I had solved this issue, but it gets complicated again. I’m using the classic chichats example as it is in the basic documentation.
That is to say…
story
checked_start
- intent x
- form_x
- form {“name”: “form_x”}
- chitchat
- respond_chitchat
- action_ask_continue
- deny
- utter_stop
- action_deactivate_form
- form {“name”: null}
intent x comes from a mapping declared in domain.yml
when the user types something corresponding to the chitchat said intent is detected with confidence = 0.99
but then in all cases the FormPolicy returns that a form is active and the chosen action is always the form again.
My config is the following:
language: is pipeline:
- name: WhitespaceTokenizer case_sensitive: false
- name: LexicalSyntacticFeaturizer
- name: CountVectorsFeaturizer analyzer: “char_wb” min_ngram: 1 max_ngram: 4
- name: DIETClassifier batch_size: [64,128] epochs: 50
- name: EntitySynonymMapper
- name: “ResponseSelector”
policies:
-
name: MemoizationPolicy max_history: 5
-
name: TEDPolicy max_history: 5 epochs: 40 state_featurizer: LabelTokenizerSingleStateFeaturizer
-
name: MappingPolicy
-
name: “FormPolicy”
-
name: “TwoStageFallbackPolicy” nlu_threshold: 0.5 ambiguity_threshold: 0.1 core_threshold: 0.5 fallback_core_action_name: “custom_default_fallback” fallback_nlu_action_name: “custom_default_fallback” deny_suggestion_intent_name: “deny_suggestion”
What is wrong or what should I change?