| CPC G10L 15/1815 (2013.01) [G06F 3/017 (2013.01); G06F 40/30 (2020.01); G06F 40/35 (2020.01); G06N 3/08 (2013.01); G06N 20/00 (2019.01); G10L 15/063 (2013.01); G10L 15/22 (2013.01); G10L 15/24 (2013.01); G10L 15/30 (2013.01); G10L 2015/227 (2013.01); G10L 2015/228 (2013.01)] | 20 Claims |

|
1. A method comprising:
obtaining training data that describes (a) a situation, (b) a gesture for a response to the situation, (c) a prompt to accompany the response, and (d) a gestural annotation for the response, the training data including a hypothetical user query related to a custom domain,
the situation being a stimulus warranting a response from a natural language understanding (NLU) model,
the gesture for the response being a suggested gesture in response to the situation from the NLU model,
the prompt to accompany the response being a suggested phrase in response to the situation from the NLU model, and
the gestural annotation for the response being a process of labeling the response to show a gestural outcome to be predicted by the NLU model;
the training data comprising an utterance, an intent, an entity, a vocabulary, a gesture, or an action usable to respond to the situation;
tagging the hypothetical user query with the gestural annotation for the response; and
training the NLU model to address the situation, based on the training data, the training comprising:
based on the tagged hypothetical user query, training the NLU model to recognize a plurality of gesture types, and
based on recognizing a gesture type of the plurality of gesture types, train the NLU model to classify and annotate a real user query at a runtime based on the recognized gesture type.
|