Speech processing dialog management

Patent metadata

Field	Value
Publication number	US-12451127-B2
Application number	US-202017099875-A
Country	US
Kind code	B2
Filing date	Nov 17, 2020
Priority date	Jan 9, 2018
Publication date	Oct 21, 2025
Grant date	Oct 21, 2025

Abstract

A system for processing user utterances and/or text based queries that tracks entities and other context data of a current dialog between the system and the user and can fill slots for new intents of the dialog by performing statistical processing on previously mentioned entities with respect to current slots to be filled. The system may compare a previously mentioned entity to a current slot to be filled using vector representations, such as word embeddings, of the current utterance, dialog history, current intent, name of an entity under consideration, category of the current slot to be filled, distance between the current dialog turn and the dialog turn that mentioned the entity, and other considerations. The individual vectors may be weighted according to an attention operation and processed by a trained decoder to output a score indicating whether the entity in consideration is relevant to the particular slot. In this manner, slots may be filled using entities from previous dialog turns, thus performing statistical anaphora resolution and leading to improved system performance.

First claim

What is claimed is:

1. A computer-implemented method, comprising:
receiving input data corresponding to a natural language user input, the input data corresponding to a user profile;
receiving dialog history data corresponding to the user profile;
determining the dialog history data is associated with a dialog session including a plurality of prior dialog exchanges, wherein a dialog exchange comprises a natural language input and corresponding system response;
determining that a first dialog exchange, of the prior dialog exchanges, is associated with the natural language user input;
determining that a second dialog exchange, of the prior dialog exchanges, is associated with the natural language user input;
determining first data representing the first dialog exchange;
determining second data representing the second dialog exchange;
processing at least the first data and the second data using an encoding component to generate first encoded data representing at least the first dialog exchange, the second dialog exchange, and an additional dialog exchange associated with the dialog history data;
identifying context data associated with the dialog history data;
processing the context data to generate second encoded data;
processing the input data, the first encoded data, and the second encoded data to generate third data;
based at least in part on the third data, selecting an application from a plurality of applications to determine a response to the natural language user input;
sending at least a portion of the input data to the application;
generating, based at least in part on a portion of the input data, response data; and
outputting the response data in response to the natural language user input.

2. The computer-implemented method of claim 1 , wherein the first encoded data represents at least one previous natural language user input corresponding to the user profile.

3. The computer-implemented method of claim 2 , wherein the first encoded data represents a recency of the at least one previous natural language user input.

4. The computer-implemented method of claim 1 , wherein the first encoded data represents at least one system response to a previous natural language user input corresponding to the user profile.

5. The computer-implemented method of claim 4 , wherein the at least one system response is associated with the application.

6. The computer-implemented method of claim 1 , wherein the first encoded data comprises encoded text data representing at least a first word corresponding to the application.

7. The computer-implemented method of claim 1 , wherein the third data represents an entity potentially corresponding to the natural language user input.

8. The computer-implemented method of claim 1 , wherein the portion of the input data comprises at least a portion of the second encoded data.

9. The computer-implemented method of claim 1 , wherein the dialog session comprises a plurality of utterances.

10. The computer-implemented method of claim 1 , wherein processing the context data includes using a second encoding component to generate the second encoded data.

11. The computer-implemented method of claim 1 , further comprising:
determining intent data associated with the input data;
processing the intent data using a third encoding component to generate third encoded data.

12. The computer-implemented method of claim 11 , wherein the third encoded data is processed to generate the third data.

13. A system comprising:
at least one processor; and
at least one memory comprising instructions that, when executed by the at least one processor, cause the system to:
receive input data corresponding to a natural language user input, the input data corresponding to a user profile;
receive dialog history data corresponding to the user profile;
determine the dialog history data is associated with a dialog session including a plurality of prior dialog exchanges, wherein a dialog exchange comprises a natural language input and corresponding system response;
determine that a first dialog exchange, of the prior dialog exchanges, is associated with the natural language user input;
determine that a second dialog exchange, of the prior dialog exchanges, is associated with the natural language user input;
determine first data representing the first dialog exchange;
determine second data representing the second dialog exchange;
process at least the first data and the second data using an encoding component to generate first encoded data representing at least the first dialog exchange, the second dialog exchange, and an additional dialog exchange association with the dialog history data;
identify context data associated with the dialog history data;
process the context data to generate second encoded data;
process the input data, the first encoded data and the second encoded data to generate third data;
based at least in part on the third data, select an application from a plurality of applications to determine a response to the natural language user input;
send at least a portion of the input data to the application;
generate, based at least in part on a portion of the input data, response data; and
output the response data in response to the natural language user input.

14. The system of claim 13 , the first encoded data represents at least one previous natural language user input corresponding to the user profile.

15. The system of claim 14 , wherein the first encoded data represents a recency of the at least one previous natural language user input.

16. The system of claim 13 , wherein the first encoded data represents at least one system response to a previous natural language user input corresponding to the user profile.

17. The system of claim 16 , wherein the at least one system response is associated with the application.

18. The system of claim 13 , wherein the first encoded data comprises encoded text data representing at least a first word corresponding to the application.

19. The system of claim 13 , wherein the third data represents an entity potentially corresponding to the natural language user input.

20. The system of claim 13 , wherein the a portion of the input data comprises at least a portion of the second encoded data.

Assignees

• Amazon Tech Inc

Inventors

• Mathias Lambert Leo
• Ummaneni Bala Murali Krishna
• Aldrich Ryan Scott
• Bishop Diamond
• Sarikaya Ruhi
• Naik Chetan Nagaraj

Classifications

• G06F16/90332 Primary

  Natural language query formulation or dialogue systems · CPC title

• G06F40/295

  Named entity recognition · CPC title

• G06F40/30

  Semantic analysis · CPC title

• G10L2015/223

  Execution procedure of a spoken command · CPC title

• G10L15/22

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