Determining intents and responses using machine learning in conversational AI systems and applications
In various examples, hybrid models for determining intents in conversational AI systems and applications are disclosed. Systems and methods are disclosed that use a machine learning model(s) and a data file(s) that associates intents with one another (e.g., using a tree-like structure) in order to determine a final intent associated with text. For example, the text may initially be processed using the machine learning model(s) (e.g., a first machine learning model) in order to determine a first intent associated with the text. The data file(s) may then be used to determine information (e.g., anchors) for one or more second intents (e.g., one or more sub-intents) that are related to the first intent. The text and the information may then be processed using the machine learning model(s) (e.g., a second machine learning model) to determine a second intent, from the one or more second intents, that is associated with the text.
1 . A method comprising:
determining, based at least on one or more machine learning models processing input data associated with a request, a first intent associated with the request, the one or more machine learning models trained to determine the first intent;
determining, using the first intent, that at least a portion of stored data represents one or more relationships between the first intent and one or more second intents which the one or more machine learning models are not trained to determine;
applying, to the one or more machine learning models, the input data and the at least the portion of the stored data;
determining, based at least on the one or more machine learning models processing the input data and the at least the portion of the stored data, a second intent of the one or more second intents that is associated with the request; and
causing, based at least on the second intent, an output of response to the request.
2 . The method of claim 1 , wherein:
the determining the first intent associated with the request uses one or more first machine learning models of the one or more machine learning models; and
the determining the second intent associated with the request uses one or more second machine learning models of the one or more machine learning models.
3 . The method of claim 1 , wherein:
the stored data is further representative of one or more second relationships between a third intent and one or more fourth intents; and
the at least the portion of the stored data does not represent the one or more second relationships between the third intent and the one or more fourth intents.
4 . The method of claim 3 , wherein the determining the second intent that is associated with the request comprises:
determining, based at least on the one or more machine learning models processing the input data and the at least the portion of the stored data, a first confidence score associated with the second intent and a second confidence score associated with a third intent of the one or more second intents; and
determining, based at least on the first confidence score being greater than the second confidence score, the second intent that is associated with the request.
5 . The method of claim 1 , wherein the stored data is representative of at least first text associated with the second intent and second text associated with the second intent.
6 . The method of claim 5 , wherein the determining the second intent that is associated with the request comprises:
determining, based at least on the one or more machine learning models processing the input data and the at least the portion of the stored data, a first confidence score associated with the first text and a second confidence score associated with the second text; and
determining, based at least on at least one of the first confidence score or the second confidence score satisfying a threshold score, the second intent that is associated with the request.
7 . The method of claim 1 , further comprising:
determining, using the one or more machine learning models and based at least on the input data and additional data associated with one or more third intents related to the second intent, one or more confidence scores associated with the one or more third intents;
determining that the one or more confidence scores do not satisfy a threshold score; and
determining, based at least on the one or more confidence scores not satisfying the threshold score, that the second intent includes a final intent associated with the request.
8 . The method of claim 1 , further comprising:
determining, using the one or more machine learning models and based at least on the input data and additional data associated with one or more third intents related to the second intent, one or more confidence scores associated with the one or more third intents;
determining that at least a confidence score of the one or more confidence scores satisfies a threshold score; and
determining, based at least on the confidence score satisfying the threshold score, that a third intent of the one or more third intents includes a final intent associated with the request,
wherein the causing the output of the response to the request is based at least on the third intent.
9 . The method of claim 1 , further comprising:
determining additional data associated with the first intent, the additional data corresponding to one or more of an action to perform a search for a new intent, one or more identifiers associated with the one or more second intents, or a confidence score associated with the first intent,
wherein the determining the second intent is further based at least on the additional data.
10 . The method of claim 1 , further comprising:
determining an action associated with the second intent, the action including at least text for generating the response; and
generating the response based at least on the text.
11 . A system comprising:
one or more processors to:
store information indicating at least one or more first relationships between a first intent and one or more second intents and one or more second relationships between the one or more second intents and one or more third intents;
determine, based at least on one or more machine learning models processing first data associated with a request, a first intent associated with the request;
determine, based at least on the one or more machine learning models processing the first intent and second data representing a first portion of the information that indicates the one or more first relationships between the first intent and the one or more second intents, a second intent of the one or more second intents;
determine, based at least on the one or more machine learning models processing the first data and third data representing a second portion of the information that indicates the one or more second relationships between the one or more second intents and the one or more third intents, a third intent of the one or more third intents; and
cause, based at least on the third intent, an output of a response to the request.
12 . The system of claim 11 , wherein the output of the response to the request is caused at least by:
determining, based at least on a confidence score associated with the third intent satisfying a threshold score, the response using the third intent; and
causing the output of the response to the request.
13 . The system of claim 11 , wherein:
the information further indicates one or more third relationships between a fourth intent and one or more fifth intents; and
the second data does not represent a third portion of the information that indicates the one or more third relationships between the fourth intent and the one or more fifth intents.
14 . The system of claim 11 , wherein:
the third intent is associated with a first confidence score;
the one or more processors are further to determine, based at least on the one or more machine learning models processing the first data and the third data, a second confidence score associated with a fourth intent of the one or more third intents; and
the response is generated using the third intent based at least on the first confidence score and the second confidence score.
15 . The system of claim 11 , wherein the system is comprised in at least one of:
a control system for an autonomous or semi-autonomous machine;
a perception system for an autonomous or semi-autonomous machine;
a system for performing simulation operations;
a system for performing digital twin operations;
a system for performing light transport simulation;
a system for performing collaborative content creation for 3D assets;
a system for performing deep learning operations;
a system implemented using an edge device;
a system implemented using a robot;
a system for performing conversational AI operations;
a system for generating synthetic data;
a system incorporating one or more virtual machines (VMs);
a system implemented at least partially in a data center; or
a system implemented at least partially using cloud computing resources.
16 . The system of claim 11 , wherein the information further indicates:
a first intent level that includes at least the first intent;
a second intent level that includes that includes at least the one or more second intents; and
a third intent level that includes at least the one or more third intents.
17 . The system of claim 11 wherein:
the one or more machine learning models are trained to determine the first intent;
the one or more machine learning models are not trained to determine the second intent; and
the second intent is determined by the one or more machine learning models based at least on the processing the first data.
18 . A processor comprising processing circuitry to:
store first data representing one or more relationships between at least first text describing a first intent that one or more machine learning models are trained to determine and second text describing a second intent that the one or more machine learning models are not trained to determine;
determine, based at least on the one or more machine learning models processing second data associated with a query, the first text describing the first intent associated with the query;
determine a portion of the first data that represents one or more of the relationships between the first text describing the first intent and the second text describing the second intent;
apply, to the one or more machine learning models, the second data and the portion of the first data that represents the one or more relationships between the first text describing the first intent and the second text describing the second intent;
determine, based at least on the one or more machine learning models processing the second data and the portion of the first data, the second text describing the second intent that is associated with the query; and
cause, based at least on the second text describing the second intent, an output of a response to the query.
19 . The processor of claim 18 , wherein the processor is comprised in at least one of:
a control system for an autonomous or semi-autonomous machine;
a perception system for an autonomous or semi-autonomous machine;
a system for performing simulation operations;
a system for performing digital twin operations;
a system for performing light transport simulation;
a system for performing collaborative content creation for 3D assets;
a system for performing deep learning operations;
a system implemented using an edge device;
a system implemented using a robot;
a system for performing conversational AI operations;
a system for generating synthetic data;
a system incorporating one or more virtual machines (VMs);
a system implemented at least partially in a data center; or
a system implemented at least partially using cloud computing resources.
20 . The method of claim 1 , further comprising:
obtaining additional data representing at least one or more instances of text that are related to the second intent; and
applying the additional data to the one or more machine learning models,
wherein the determining the second intent that is associated with the request is further based at least on the one or more machine learning models processing the additional data.