Interpreting voice commands

Patent metadata

Field	Value
Publication number	US-9972318-B1
Application number	US-201615271758-A
Country	US
Kind code	B1
Filing date	Sep 21, 2016
Priority date	Sep 21, 2016
Publication date	May 15, 2018
Grant date	May 15, 2018

Abstract

A system capable of generating and storing progress data associated with third party services. A voice enabled device may receive voice commands and the system may perform natural language understanding (NLU) to interpret the voice commands, determine a corresponding process and send instructions to a third party server associated with the process. In order to resume the process after the process is interrupted, the system may save and/or cause the third party server to save progress data and/or a checkpoint corresponding to a status of the process at the time that the process is interrupted. The system may use the progress data to resume the process at a later point based on the status. In addition, the system may track processes associated with saved progress data/checkpoints and may interpret incoming voice commands based on the tracked processes.

First claim

What is claimed is:

1. A computer-implemented method for interpreting a voice command for a speech application, the method comprising:
receiving, from a first device during a first time period, audio data corresponding to an utterance;
performing speech recognition on the audio data to determine text data;
determining a first command corresponding to the text data;
determining a first word included in the first command;
determining a list of applications that are associated with progress data, the list of applications including a first application;
determining, using an association table corresponding to the list of applications, that the first word is associated with the first application;
determining, using a first model corresponding to a first domain, a confidence score that the first command is associated with the first application;
determining that the confidence score is above a threshold;
determining, using the association table, first progress data associated with the first application, wherein the first progress data includes information about a first status of the first application during a previous session associated with the first application; and
causing the first application to resume using the first progress data.

2. The computer-implemented method of claim 1 , further comprising:
determining, during a second time period before the first time period began, to halt the previous session associated with the first application;
detecting an indicator indicating that the first application is capable of using progress data to represent a current status of the first application;
determining the first progress data corresponding to the first application, wherein the first progress data includes information about the first status of the first application during the second time period, the first status corresponding to a current step in a series of steps associated with the first application;
storing an association between the first progress data and an identity of the first application in the association table;
adding the identity of the first application to the list of applications; and
halting, during the second time period, the previous session associated with the first application.

3. The computer-implemented method of claim 1 , wherein determining that the first word is associated with the first application further comprises:
identifying the first word within the association table, the association table indicating at least two of an identity of the first application, an active task associated with the first application, a task type of the active task, a session identification associated with the previous session, keywords corresponding to the first application and information corresponding to the first progress data; and
determining that the first word is included in the keywords corresponding to the first application.

4. The computer-implemented method of claim 1 , further comprising:
processing the text data with a second model to determine:
a second command, and
a second confidence score corresponding to the second command;

processing the text data with the first model to determine:
the first command, and
a third confidence score corresponding to the third command;

determining that the third confidence score is greater than the second confidence score;
determining that the first command corresponds to the text data; and
determining, based on the second model, the confidence score that the first command is associated with the first application.

5. A computer-implemented method, comprising:
receiving, from a first device during a first time period, audio data corresponding to an utterance;
performing speech recognition on the audio data to determine text data;
determining a first command corresponding to the text data;
determining a confidence score that the first command is associated with a first process, the first process included in a list of processes associated with progress data;
determining that the confidence score is above a threshold;
determining first progress data associated with the first process, wherein the first progress data includes information about a first status of the first process during a previous session associated with the first process; and
causing the first process to resume based on the first progress data.

6. The computer-implemented method of claim 5 , further comprising:
determining, during a second time period before the first time period began, to halt the previous session associated with the first process;
determining that the first process is capable of using progress data to represent a current status of the first process;
determining the first progress data corresponding to the first process, wherein the first progress data includes information about the first status of the first process during the second time period;
storing an association between the first progress data and an identity of the first process in a user profile associated with the first device;
adding the identity of the first process to the list of processes associated with progress data; and
halting, during the second time period, the previous session associated with the first process.

7. The computer-implemented method of claim 6 , further comprising:
sending, during the second time period to a second device associated with the first process, a first request to create and store second progress data that corresponds to the first progress data;
sending, to the second device during the first time period, a second request to resume operating the first process based on the second progress data, the second request including the first progress data, the first progress data comprising at least one of a reference location or a current step of a series of steps associated with the first process.

8. The computer-implemented method of claim 6 , further comprising:
sending, during the second time period to a second device associated with the first process, a first request to create and send the first progress data;
receiving the first progress data from the second device;
storing the first progress data; and
sending, to the second device during the first time period, a second request to resume operating the first process based on the first progress data, the second request including the first progress data.

9. The computer-implemented method of claim 5 , further comprising:
processing the text data with a first model to determine:
a second command, and
a second confidence score corresponding to the second command;

processing the text data with a second model to determine:
the first command, and
a third confidence score corresponding to the third command;

determining that the third confidence score is greater than the second confidence score;
determining that the first command corresponds to the text data; and
determining, based on the second model, the confidence score that the first command is associated with the first process.

10. The computer-implemented method of claim 5 , further comprising:
performing the speech recognition on the audio data using a first model to determine the text data;
determining a second confidence score that the text data corresponds to the audio data;
performing second speech recognition on the audio data using a second model to determine second text data;
determining a third confidence score that the second text data corresponds to the audio data;
determining that the second confidence score is greater than the third confidence score;
determining, based on the first model, the first command corresponding to the text data; and
determining, based on the first model, the confidence sc

Assignees

• Amazon Tech Inc

Inventors

• Kelly Eric
• Utter Kevin

Classifications

• G10L15/22 Primary

  Procedures used during a speech recognition process, e.g. man-machine dialogue · CPC title

• G10L15/265

  Physics · mapped topic

• G10L15/222 Primary

  Barge in, i.e. overridable guidance for interrupting prompts · CPC title

• G10L15/30

  Distributed recognition, e.g. in client-server systems, for mobile phones or network applications · CPC title

• G10L15/08

  Speech classification or search · CPC title