notice
This is unreleased documentation for Rasa Documentation Main/Unreleased version.
For the latest released documentation, see the latest version (3.x).
rasa.core.policies.unexpected_intent_policy
RankingCandidateMetadata Objects
Dataclass to represent metada for a candidate intent.
UnexpecTEDIntentPolicyMetadata Objects
Dataclass to represent policy metadata.
UnexpecTEDIntentPolicy Objects
UnexpecTEDIntentPolicy
has the same model architecture as TEDPolicy
.
The difference is at a task level. Instead of predicting the next probable action, this policy predicts whether the last predicted intent is a likely intent according to the training stories and conversation context.
get_default_config
Returns the default config (see parent class for full docstring).
__init__
Declares instance variables with default values.
model_class
Gets the class of the model architecture to be used by the policy.
Returns:
Required class.
compute_label_quantiles_post_training
Computes quantile scores for prediction of action_unlikely_intent
.
Multiple quantiles are computed for each label
so that an appropriate threshold can be picked at
inference time according to the tolerance
value specified.
Arguments:
model_data
- Data used for training the model.label_ids
- Numerical IDs of labels for each data point used during training.
run_training
Feeds the featurized training data to the model.
Arguments:
model_data
- Featurized training data.label_ids
- Label ids corresponding to the data points inmodel_data
.
Raises:
RasaCoreException
if label_ids
is None as it's needed for
running post training procedures.
predict_action_probabilities
Predicts the next action the bot should take after seeing the tracker.
Arguments:
tracker
- Tracker containing past conversation events.domain
- Domain of the assistant.rule_only_data
- Slots and loops which are specific to rules and hence should be ignored by this policy.precomputations
- Contains precomputed features and attributes.
Returns:
The policy's prediction (e.g. the probabilities for the actions).
persist_model_utilities
Persists model's utility attributes like model weights, etc.
Arguments:
model_path
- Path where model is to be persisted
IntentTED Objects
Follows TED's model architecture from https://arxiv.org/abs/1910.00486.
However, it has been re-purposed to predict multiple labels (intents) instead of a single label (action).
dot_product_loss_layer
Returns the dot-product loss layer to use.
Multiple intents can be valid simultaneously, so IntentTED
uses the
MultiLabelDotProductLoss
.
Returns:
The loss layer that is used by _prepare_dot_product_loss
.
run_bulk_inference
Computes model's predictions for input data.
Arguments:
model_data
- Data to be passed as input
Returns:
Predictions for the input data.