# US 2024/0203412 A1 · Enterprise type models for voice interfaces

> Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt

- Number: US 2024/0203412 A1
- Document ID: US20240203412A1
- Status: Published application
- Kind: A1
- Application number: 18/082,786
- Filed: 2022-12-16
- Published: 2024-06-20
- Earliest claimed: 2022-12-16
- Assignees: Amazon Technologies, Inc.
- Applicants: Amazon Technologies, Inc.
- Inventors: Austin Chang Ming Liu; Gonzalo Alvarez Barrio; Gregory James Wade; Harsh Agarwal; Sam Anthony Sullivan; Chieh Chien; Cameron L Chinn
- Web page: https://patentgrep.com/patent/US20240203412A1

## CPC

- G10L 15/183: using context dependencies, e.g. language models
- G10L 15/22: Procedures used during a speech recognition process, e.g. man-machine dialogue
- G10L 15/063: {Training}
- G10L: SPEECH ANALYSIS TECHNIQUES OR SPEECH SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING TECHNIQUES; SPEECH OR AUDIO CODING OR DECODING

## Abstract

Systems and methods for enterprise type pretrained models for voice interfaces include the generation and validation of enterprise type pretrained models utilizing input associated with the enterprise type at issue. Once generated and validated, when a user command is received, the speech processing system may check to determine if a customized model is available, and if not, may query the enterprise type model to provide a response to the user command.

## Claims

### Claim 1

A system, comprising:
one or more processors; and
non-transitory computer-readable media storing computer-executable instructions that, when executed by the one or more processors, cause the one or more processors to perform operations comprising:
receiving, from a voice interface device, audio data representing a user utterance including a request that an action be performed, the voice interface device associated with account data identifying the voice interface device as being associated with an enterprise;
determining that the account data is associated with an enterprise type model configured to identify a plurality of enterprise applications to service user requests, wherein the enterprise type model is generated based on an enterprise type associated with the enterprise, the enterprise type indicating an industry of the enterprise;
retrieving the enterprise type model from a model repository;
applying the enterprise type model to the user utterance to identify a first enterprise application from the plurality of enterprise applications;
ranking the first enterprise application with other candidate applications to service the request;
selecting the first enterprise application to service the request;
generating, utilizing the first enterprise application, a speech processing result associated with the action to be performed; and
sending, to the voice interface device, a directive to cause the action to be performed.

### Claim 2 (depends on 1)

The system of claim 1, the operations further comprising:
receiving first data indicating the enterprise type associated with the enterprise;
determining a subset of intents associated with the enterprise type, wherein the subset of intents represents a portion of potential intents that have been predefined to be associated with the enterprise type; and
generating the enterprise type model such that the subset of intents are identified as being serviceable by the first enterprise application.

### Claim 3 (depends on 1)

The system of claim 1, the operations further comprising:
receiving, from a device associated with the enterprise, first input data requesting generation of the enterprise type model in association with the voice interface device;
causing display of a user interface configured to receive second input data indicating the enterprise type of the enterprise;
receiving the second input data;
in response to the second input data, causing the user interface to display voice command options predefined to be associated with the enterprise type;
receiving third input data representing selection of one or more of the voice command options; and
generating the enterprise type model utilizing the third input data.

### Claim 4 (depends on 1)

The system of claim 1, the operations further comprising:
determining that the audio data was received from the voice interface device; and
in response to the audio data being received from the voice interface device, selecting the enterprise type model to be utilized to identify the first enterprise type model.

### Claim 5

A method, comprising:
receiving first input data requesting that an action be performed from a first device associated with an enterprise;
based at least in part on the first device being associated with the enterprise and account data associated with the enterprise indicating availability of an enterprise type model to determine an application to cause the action to be performed, selecting the application to cause the action to be performed;
generating, utilizing the application, a first result associated with the action to be performed; and
causing the action to be performed utilizing the first result.

### Claim 6 (depends on 5)

The method of claim 5, further comprising:
receiving first data indicating an enterprise type associated with the enterprise;
determining a subset of inputs to associate with the enterprise type model based at least in part on the enterprise type; and
generating the enterprise type model such that the subset of inputs are identified as being associated with the application.

### Claim 7 (depends on 5)

The method of claim 5, further comprising:
receiving, from a second device associated with the enterprise, second input data requesting generation of the enterprise type model in association with the first device;
causing display of a user interface configured to receive third input data indicating the enterprise type of the enterprise;
receiving the third input data;
based at least in part on the third input data, causing the user interface to display user command response options predefined to be associated with the enterprise type;
receiving fourth input data representing selection of one or more of the user command response options; and
generating the enterprise type model utilizing the fourth input data.

### Claim 8 (depends on 5)

The method of claim 5, further comprising:
determining that the first input data was received from the first device; and
based at least in part on the first input data being received from the first device, determining that the enterprise type model is available in a model repository for selecting the application.

### Claim 9 (depends on 5)

The method of claim 5, further comprising:
determining text data based at least in part on the first input data; and
determining, utilizing the enterprise type model, that the text data is associated with an enterprise-specific request instead of an enterprise-agnostic request, wherein selecting the application is based at least in part on the text data being associated with the enterprise-specific request.

### Claim 10 (depends on 5)

The method of claim 5, further comprising:
receiving second input data requesting that the action be performed;
determining that the second input data includes an enterprise-agnostic request; and
based at least in part on the second input data including an enterprise-agnostic request and utilizing the enterprise type model, determining to select an enterprise-agnostic domain to service the enterprise-agnostic request.

### Claim 11 (depends on 5)

The method of claim 5, further comprising:
generating a first version of the enterprise type model configured to be utilized for selecting candidate applications for responding to voice commands received at a microphone of the first device;
generating a second version of the enterprise type model configured to be utilized for selecting the candidate applications for responding to touch input received at a screen of the first device;
determining that the first input data is a voice command instead of the touch input; and
utilizing the first version of the enterprise type model based at least in part on the first input data being the voice command.

### Claim 12 (depends on 5)

The method of claim 5, further comprising:
determining a subset of prior use data associated with an enterprise type of the enterprise;
generating training data from the subset of the prior use data; and
training the enterprise type model utilizing the training data.

### Claim 13

A system, comprising:
one or more processors; and
non-transitory computer-readable media storing computer-executable instructions that, when executed by the one or more processors, cause the one or more processors to perform operations comprising:
receiving first input data requesting that an action be performed from a first device associated with an enterprise;
based at least in part on the first device being associated with the enterprise and account data associated with the enterprise indicating availability of an enterprise type model to determine an application to cause the action to be performed, selecting the application to cause the action to be performed;
generating, utilizing the application, a first result associated with the action to be performed; and
causing the action to be performed utilizing the first result.

### Claim 14 (depends on 13)

The system of claim 13, the operations further comprising:
receiving first data indicating an enterprise type associated with the enterprise;
determining a subset of inputs to associate with the enterprise type model based at least in part on the enterprise type; and
generating the enterprise type model such that the subset of inputs are identified as being associated with the application.

### Claim 15 (depends on 13)

The system of claim 13, the operations further comprising:
receiving, from a second device associated with the enterprise, second input data requesting generation of the enterprise type model in association with the first device;
causing display of a user interface configured to receive third input data indicating the enterprise type of the enterprise;
receiving the third input data;
based at least in part on the third input data, causing the user interface to display user command response options predefined to be associated with the enterprise type;
receiving fourth input data representing selection of one or more of the user command response options; and
generating the enterprise type model utilizing the fourth input data.

### Claim 16 (depends on 13)

The system of claim 13, the operations further comprising:
determining that the first input data was received from the first device; and
based at least in part on the first input data being received from the first device, determining that the enterprise type model is available in a model repository for selecting the application.

### Claim 17 (depends on 13)

The system of claim 13, the operations further comprising:
determining text data based at least in part on the first input data; and
determining, utilizing the enterprise type model, that the text data is associated with an enterprise-specific request instead of an enterprise-agnostic request, wherein selecting the application is based at least in part on the text data being associated with the enterprise-specific request.

### Claim 18 (depends on 13)

The system of claim 13, the operations further comprising:
receiving second input data requesting that the action be performed;
determining that the second input data includes an enterprise-agnostic request; and
based at least in part on the second input data including an enterprise-agnostic request and utilizing the enterprise type model, determining to select an enterprise-agnostic domain to service the enterprise-agnostic request.

### Claim 19 (depends on 13)

The system of claim 13, the operations further comprising:
generating a first version of the enterprise type model configured to be utilized for selecting candidate applications for responding to voice commands received at a microphone of the first device;
generating a second version of the enterprise type model configured to be utilized for selecting the candidate applications for responding to touch input received at a screen of the first device;
determining that the first input data is a voice command instead of the touch input; and
utilizing the first version of the enterprise type model based at least in part on the first input data being the voice command.

### Claim 20 (depends on 13)

The system of claim 13, the operations further comprising:
determining a subset of prior use data associated with an enterprise type of the enterprise;
generating training data from the subset of the prior use data; and
training the enterprise type model utilizing the training data.
