Speech recognition entity resolution

• US 10,366,690 B1
• Filed: 05/15/2017
• Issued: 07/30/2019
• Est. Priority Date: 05/15/2017
• Status: Active Grant

• Current Assignee: Amazon Technologies, Inc. (Amazon.com, Inc.)
• Original Assignee: Amazon Technologies, Inc. (Amazon.com, Inc.)
• Inventors: Sarikaya, Ruhi
• Primary Examiner(s): Riley, Marcus T
• Application Number:
  US15/595,466
• Time in Patent Office: 806 Days
• Field of Search: None
• CPC Class Codes: G06F 16/433 using audio data
• CPC Class Codes: G06F 3/167 Audio in a user interface, e.g. using voice commands for navigating, audio feedback
• CPC Class Codes: G06F 40/295 Named entity recognition
• CPC Class Codes: G06F 40/30 Semantic analysis
• CPC Class Codes: G10H 5/005 Voice controlled instruments
• CPC Class Codes: G10L 15/22 Procedures used during a speech recognition process, e.g. man-machine dialogue
• CPC Class Codes: G10L 2015/223 Execution procedure of a spoken command
• CPC Class Codes: G10L 25/54 for retrieval
• CPC Class Codes: G10L 25/69 for evaluating synthetic or decoded voice signals

Abstract

A canonicalizer can be used with or implemented within a natural language understanding (NLU) component of a speech processing system to enable the system to properly identify an entity to which a user refers in a spoken utterance. The canonicalizer may determine a first canonical value using a gazetteer associated with a determined intent of the spoken utterance. The canonicalizer may determine a second canonical value using a look-up table of canonical values. The canonicalizer may then output either the first canonical value or the second canonical value for further NLU processing, such as entity resolution.

Claims

1. A computing system, comprising:

at least one processor; and

at least one memory including instructions that, when executed by the at least one processor, cause the computing system to;

receive text data corresponding to a command executable by the computing system;

determine, based at least in part on the text data, an intent corresponding to the command;

determine first text string data of the text data, the first text string data corresponding to an entity;

determine a lexicon associated with the intent;

determine, in the lexicon, second text string data corresponding to the entity, wherein the second text string data is different from the first text string data and the second text string data corresponds to a recognized format for referring to the entity with respect to the intent; and

cause an action to be executed for the intent using the second text string data.

2. The computing system of claim 1, wherein the at least one memory further includes instructions that, when executed by the at least one processor, further cause the computing system to:

perform, with respect to the first text string data, at least one of case normalization, punctuation addition, punctuation removal, white space addition, white space removal, word level substring matching, phonetic level substring matching, number-to-text mapping, text-to-number matching, or acronym generation.

3. The computing system of claim 1, wherein the at least one memory further includes instructions that, when executed by the at least one processor, further cause the computing system to:

determine, based at least in part on the text data, a second intent potentially corresponding to the command;

determine a second lexicon associated with the second intent;

determine, in the second lexicon, third text string data corresponding to the entity, wherein the third text string data is different from the first text string data; and

cause one of the action to be executed for the intent using the second text string data or a second action to be executed for the second intent using the third text string data.

4. The computing system of claim 1, wherein the at least one memory further includes instructions that, when executed by the at least one processor, further cause the computing system to:

determine an entity type corresponding to the entity; and

determine, in the lexicon, the second text string data corresponding to the entity type.

5. The computing system of claim 1, wherein the second text string data includes the first text string data.

6. The computing system of claim 1, wherein the at least one memory further includes instructions that, when executed by the at least one processor, further cause the computing system to:

append the second text string data to the first text string data.

7. The computing system of claim 1, wherein the at least one memory further includes instructions that, when executed by the at least one processor, further cause the computing system to:

using a trained model, process the first text string data, a first indicator corresponding to the intent, and a second indicator corresponding to an entity list to determine the second text string data.

8. The computing system of claim 1, wherein the at least one memory further includes instructions that, when executed by the at least one processor, further cause the computing system to:

receive, from an application developer device, word string data;

store the word string data in a look-up table associated with a first application;

determine the first text string data is represented in third text string data included in the look-up table; and

cause the action to be executed using the third text string data.

9. A computer-implemented method, comprising:

receiving text data corresponding to a command executable by a computing system;

determining, based at least in part on the text data, an intent corresponding to the command;

determining first text string data of the text data, the first text string data corresponding to an entity;

determining a lexicon associated with the intent;

determining, in the lexicon, second text string data corresponding to the entity, wherein the second text string data is different from the first text string data and the second text string data corresponds to a recognized format for referring to the entity with respect to the intent; and

causing an action to be executed for the intent using the second text string data.

10. The computer-implemented method of claim 9, wherein determining the second text string data comprises:

performing, with respect to the first text string data, at least one of case normalization, punctuation addition, punctuation removal, white space addition, white space removal, word level substring matching, phonetic level substring matching, number-to-text mapping, text-to-number matching, or acronym generation.

11. The computer-implemented method of claim 9, further comprising:

determining, based at least in part on the text data, a second intent potentially corresponding to the command;

determining a second lexicon associated with the second intent;

determining, in the second lexicon, third text string data corresponding to the entity, wherein the third text string data is different from the first text string data; and

causing one of the action to be executed for the intent using the second text string data or a second action to be executed for the second intent using the third text string data.

12. The computer-implemented method of claim 9, further comprising:

determining an entity type corresponding to the entity; and

determining, in the lexicon, the second text string data corresponds to the entity type.

13. The computer-implemented method of claim 9, wherein the second text string data includes the first text string data.

14. The computer-implemented method of claim 9, further comprising:

appending the second text string data to the first text string data.

15. The computer-implemented method of claim 9, further comprising:

using a trained model, processing the first text string data, a first indicator corresponding to the intent, and a second indicator correspond to an entity list to determine the second text string data.

16. The computer-implemented method of claim 9, further comprising:

receiving, from an application developer device, word string data;

storing the word string data in a look-up table associated with a first application;

determining that the first text string data is represented in third text string data included in the look-up table; and

causing the action to be executed using the third text string data.

17. A computing system, comprising:

at least one processor; and

at least one memory including instructions that, when executed by the at least one processor, cause the computing system to;

receive text data corresponding to a command executable by the computing system;

determine, based at least in part on the text data, an intent corresponding to the command;

determine first text string data of the text data, the first text string data corresponding to an entity;

determine a lexicon associated with the intent;

determine, in the lexicon, second text string data corresponding to the entity, wherein the second text string data is different from the first text string data;

receive, from an application developer device, word string data;

store the word string data in a look-up table associated with a first application; and

cause an action to be executed for the intent using the second text string data and third text string data.

18. The computing system of claim 17, wherein the at least one memory further includes instructions that, when executed by the at least one processor, further cause the computing system to:

perform, with respect to the first text string data, at least one of case normalization, punctuation addition, punctuation removal, white space addition, white space removal, word level substring matching, phonetic level substring matching, number-to-text mapping, text-to-number matching, or acronym generation.

19. The computing system of claim 17, wherein the at least one memory further includes instructions that, when executed by the at least one processor, further cause the computing system to:

determine an entity type corresponding to the entity; and

determine, in the lexicon, the second text string data corresponding to the entity type.

20. The computing system of claim 17, wherein the at least one memory further includes instructions that, when executed by the at least one processor, further cause the computing system to:

append the second text string data to the first text string data.