Hybrid local and global processing of voice requests from users
Systems and methods described herein may provide a system that enables hybrid local and global processing of voice requests from a user. A computing device may receive audio data containing a voice request from a user and may determine, using a local intent engine, a local intent prediction based on the voice request. The computing device may determine that the local intent prediction has a confidence measure that is less than a predetermined threshold and may, in response, transmit data containing the voice request to a global intent engine. The computing device may receive a global intent prediction from the global intent engine and may determine a corresponding computing process based on the global intent prediction.
1 . A method, comprising:
receiving, at a first computing device, audio data containing a voice request from a user;
determining, using a local intent engine executing on the first computing device, a local intent prediction based on the voice request;
determining that a first confidence measure for the local intent prediction is less than a predetermined threshold;
transmitting data containing the voice request to a global intent engine executing on a second computing device;
receiving, from the global intent engine, a global intent prediction based on the voice request, wherein the global intent prediction includes (i) a second predicted intent of the user indicated by the voice request and (ii) a second confidence measure for the second predicted intent;
determining, based on the global intent prediction, a corresponding computing process, wherein determining the corresponding computing process further comprises:
determining that the second confidence measure is greater than or equal to a predetermined threshold; and
identifying the corresponding computing process as corresponding to the second predicted intent; and
executing the corresponding computing process.
2 . The method of claim 1 , wherein the local intent prediction includes a first predicted intent of the user indicated by the voice request, and wherein the first confidence measure is for the first predicted intent.
3 . The method of claim 1 ,
wherein the data containing the voice request is transmitted to the global intent engine in response to determining that the first confidence measure is less than the predetermined threshold.
4 . The method of claim 1 , wherein determining the corresponding computing process further comprises:
determining that the second confidence measure is less than a predetermined threshold; and
requesting additional user input to indicate the corresponding computing process.
5 . The method of claim 4 , further comprising updating, based on the additional user input, at least one of the local intent engine and the global intent engine.
6 . The method of claim 1 , further comprising, before determining the local intent prediction:
generating, based on the audio data, a text containing the voice request,
wherein the text is provided to the local intent engine to determine the local intent prediction.
7 . The method of claim 1 , wherein the local intent engine is a machine learning model trained based on previous requests received by the first computing device from the user.
8 . An information handling system, comprising:
a memory; and
a processor coupled to the memory, wherein the processor is configured to perform steps comprising:
receiving, at a first computing device, audio data containing a voice request from a user;
determining, using a local intent engine executing on the first computing device, a local intent prediction based on the voice request;
determining that a first confidence measure for the local intent prediction is less than a predetermined threshold;
transmitting data containing the voice request to a global intent engine executing on a second computing device;
receiving, from the global intent engine, a global intent prediction based on the voice request, wherein the global intent prediction includes (i) a second predicted intent of the user indicated by the voice request and (ii) a second confidence measure for the second predicted intent;
determining, based on the global intent prediction, a corresponding computing process, wherein determining the corresponding computing process further comprises:
determining that the second confidence measure is greater than or equal to a predetermined threshold; and
identifying the corresponding computing process as corresponding to the second predicted intent; and
executing the corresponding computing process.
9 . The information handling system of claim 8 , wherein the local intent prediction includes a first predicted intent of the user indicated by the voice request, and wherein the first confidence measure is for the first predicted intent.
10 . The information handling system of claim 8 ,
wherein the data containing the voice request is transmitted to the global intent engine in response to determining that the first confidence measure is less than the predetermined threshold.
11 . The information handling system of claim 8 , wherein determining the corresponding computing process further comprises:
determining that the second confidence measure is less than a predetermined threshold; and
requesting additional user input to indicate the corresponding computing process.
12 . A computer program product, comprising:
a non-transitory computer readable medium comprising code which, when executed by a processor, causes the processor to perform steps comprising:
receiving, at a first computing device, audio data containing a voice request from a user;
determining, using a local intent engine executing on the first computing device, a local intent prediction based on the voice request;
determining that a first confidence measure for the local intent prediction is less than a predetermined threshold;
transmitting data containing the voice request to a global intent engine executing on a second computing device;
receiving, from the global intent engine, a global intent prediction based on the voice request, wherein the global intent prediction includes (i) a second predicted intent of the user indicated by the voice request and (ii) a second confidence measure for the second predicted intent;
determining, based on the global intent prediction, a corresponding computing process, wherein determining the corresponding computing process further comprises:
determining that the second confidence measure is greater than or equal to a predetermined threshold; and
identifying the corresponding computing process as corresponding to the second predicted intent; and
executing the corresponding computing process.
13 . The computer program product of claim 12 , wherein the local intent prediction includes a first predicted intent of the user indicated by the voice request, and wherein the first confidence measure is for the first predicted intent.
14 . The computer program product of claim 12 ,
wherein the data containing the voice request is transmitted to the global intent engine in response to determining that the first confidence measure is less than the predetermined threshold.