This is documentation for Rasa Open Source Documentation v2.2.x, which is no longer actively maintained.
For up-to-date documentation, see the latest version (2.3.x).

Version: 2.2.x



featurize_training_examples(training_examples: List[Message], attributes: List[Text], entity_tag_specs: Optional[List["EntityTagSpec"]] = None, featurizers: Optional[List[Text]] = None, bilou_tagging: bool = False) -> List[Dict[Text, List["Features"]]]

Converts training data into a list of attribute to features.

Possible attributes are, for example, INTENT, RESPONSE, TEXT, ACTION_TEXT, ACTION_NAME or ENTITIES.


  • training_examples - the list of training examples
  • attributes - the attributes to consider
  • entity_tag_specs - the entity specs
  • featurizers - the featurizers to consider
  • bilou_tagging - indicates whether BILOU tagging should be used or not


A list of attribute to features.


convert_to_data_format(features: Union[
List[List[Dict[Text, List["Features"]]]], List[Dict[Text, List["Features"]]]
], fake_features: Optional[Dict[Text, List["Features"]]] = None, consider_dialogue_dimension: bool = True, featurizers: Optional[List[Text]] = None) -> Tuple[Data, Optional[Dict[Text, List["Features"]]]]

Converts the input into "Data" format.

"features" can, for example, be a dictionary of attributes (INTENT, TEXT, ACTION_NAME, ACTION_TEXT, ENTITIES, SLOTS, FORM) to a list of features for all dialogue turns in all training trackers. For NLU training it would just be a dictionary of attributes (either INTENT or RESPONSE, TEXT, and potentially ENTITIES) to a list of features for all training examples.

The "Data" format corresponds to Dict[Text, Dict[Text, List[FeatureArray]]]. It's a dictionary of attributes (e.g. TEXT) to a dictionary of secondary attributes (e.g. SEQUENCE or SENTENCE) to the list of actual features.


  • features - a dictionary of attributes to a list of features for all examples in the training data
  • fake_features - Contains default feature values for attributes
  • consider_dialogue_dimension - If set to false the dialogue dimension will be removed from the resulting sequence features.
  • featurizers - the featurizers to consider


Input in "Data" format and fake features