Is there way control DIET classifier to pick only entities which are specified in Domain file?

Rasa Open Source
1

DIET classifier is automatically making new entities rather than picking the entities defined in the domain file. Due to that my testing is not being successful.

In this case, it is picking “maternity” perfectly fine.

input:- I want to know about maternity leave

In the below phrase, I haven’t specified “leaves” as an entity, but DIET is picking “leaves” as an entity. input:- my wife and I are planning on adopting a child and can you tell me about leaves`

My config file is:-

  • name: WhitespaceTokenizer case_sensitive: False
  • name: RegexFeaturizer
  • name: LexicalSyntacticFeaturizer
  • name: CountVectorsFeaturizer
  • name: CountVectorsFeaturizer analyzer: char_wb min_ngram: 1 max_ngram: 4
  • name: DIETClassifier epochs: 100
  • name: EntitySynonymMapper
2

Hmm, it’s a little hard to be certain without seeing your NLU file, but it could be that “leaves” is being picked up as a synonym by the EntitySynonymMapper. Do you have any entity training examples with “leaves” or “leave” in them?