notice

This is documentation for Rasa Documentation v2.x, which is no longer actively maintained.
For up-to-date documentation, see the latest version (3.x).

Version: 2.x

rasa.core.featurizers

SingleStateFeaturizer Objects

class SingleStateFeaturizer()

Base class for mechanisms to transform the conversations state into ML formats.

Subclasses of SingleStateFeaturizer decide how the bot will transform the conversation state to a format which a classifier can read: feature vector.

prepare_from_domain

| prepare_from_domain(domain: Domain) -> None

Helper method to init based on domain.

encode

| encode(state: Dict[Text, float]) -> np.ndarray

Encode user input.

action_as_one_hot

| @staticmethod
| action_as_one_hot(action: Text, domain: Domain) -> np.ndarray

Encode system action as one-hot vector.

create_encoded_all_actions

| create_encoded_all_actions(domain: Domain) -> np.ndarray

Create matrix with all actions from domain encoded in rows.

BinarySingleStateFeaturizer Objects

class BinarySingleStateFeaturizer(SingleStateFeaturizer)

Assumes all features are binary.

All features should be either on or off, denoting them with 1 or 0.

__init__

| __init__() -> None

Declares instant variables.

prepare_from_domain

| prepare_from_domain(domain: Domain) -> None

Use Domain to prepare featurizer.

encode

| encode(state: Dict[Text, float]) -> np.ndarray

Returns a binary vector indicating which features are active.

Given a dictionary of states (e.g. 'intent_greet', 'prev_action_listen',...) return a binary vector indicating which features of self.input_features are in the bag. NB it's a regular double precision float array type.

For example with two active features out of five possible features this would return a vector like [0 0 1 0 1]

If intent features are given with a probability, for example with two active features and two uncertain intents out of five possible features this would return a vector like [0.3, 0.7, 1.0, 0, 1.0].

If this is just a padding vector we set all values to -1. padding vectors are specified by a None or [None] value for states.

create_encoded_all_actions

| create_encoded_all_actions(domain: Domain) -> np.ndarray

Create matrix with all actions from domain encoded in rows as bag of words

LabelTokenizerSingleStateFeaturizer Objects

class LabelTokenizerSingleStateFeaturizer(SingleStateFeaturizer)

Creates bag-of-words feature vectors.

User intents and bot action names are split into tokens and used to create bag-of-words feature vectors.

Arguments:

  • split_symbol - The symbol that separates words in intets and action names.
  • use_shared_vocab - The flag that specifies if to create the same vocabulary for user intents and bot actions.

__init__

| __init__(use_shared_vocab: bool = False, split_symbol: Text = "_") -> None

inits vocabulary for label bag of words representation

prepare_from_domain

| prepare_from_domain(domain: Domain) -> None

Creates internal vocabularies for user intents and bot actions.

encode

| encode(state: Dict[Text, float]) -> np.ndarray

Returns a binary vector indicating which tokens are present.

create_encoded_all_actions

| create_encoded_all_actions(domain: Domain) -> np.ndarray

Create matrix with all actions from domain encoded in rows as bag of words

TrackerFeaturizer Objects

class TrackerFeaturizer()

Base class for actual tracker featurizers.

training_states_and_actions

| training_states_and_actions(trackers: List[DialogueStateTracker], domain: Domain) -> Tuple[List[List[Dict]], List[List[Text]]]

Transforms list of trackers to lists of states and actions.

featurize_trackers

| featurize_trackers(trackers: List[DialogueStateTracker], domain: Domain) -> DialogueTrainingData

Create training data.

prediction_states

| prediction_states(trackers: List[DialogueStateTracker], domain: Domain) -> List[List[Dict[Text, float]]]

Transforms list of trackers to lists of states for prediction.

create_X

| create_X(trackers: List[DialogueStateTracker], domain: Domain) -> np.ndarray

Create X for prediction.

load

| @staticmethod
| load(path) -> Optional["TrackerFeaturizer"]

Loads the featurizer from file.

FullDialogueTrackerFeaturizer Objects

class FullDialogueTrackerFeaturizer(TrackerFeaturizer)

Creates full dialogue training data for time distributed architectures.

Creates training data that uses each time output for prediction. Training data is padded up to the length of the longest dialogue with -1.

training_states_and_actions

| training_states_and_actions(trackers: List[DialogueStateTracker], domain: Domain) -> Tuple[List[List[Dict]], List[List[Text]]]

Transforms list of trackers to lists of states and actions.

Training data is padded up to the length of the longest dialogue with -1.

prediction_states

| prediction_states(trackers: List[DialogueStateTracker], domain: Domain) -> List[List[Dict[Text, float]]]

Transforms list of trackers to lists of states for prediction.

MaxHistoryTrackerFeaturizer Objects

class MaxHistoryTrackerFeaturizer(TrackerFeaturizer)

Slices the tracker history into max_history batches.

Creates training data that uses last output for prediction. Training data is padded up to the max_history with -1.

slice_state_history

| @staticmethod
| slice_state_history(states: List[Dict[Text, float]], slice_length: Optional[int]) -> List[Optional[Dict[Text, float]]]

Slices states from the trackers history.

If the slice is at the array borders, padding will be added to ensure the slice length.

training_states_and_actions

| training_states_and_actions(trackers: List[DialogueStateTracker], domain: Domain) -> Tuple[List[List[Optional[Dict[Text, float]]]], List[List[Text]]]

Transforms list of trackers to lists of states and actions.

Training data is padded up to the max_history with -1.

prediction_states

| prediction_states(trackers: List[DialogueStateTracker], domain: Domain) -> List[List[Dict[Text, float]]]

Transforms list of trackers to lists of states for prediction.