System and method of spoken language understanding in human computer dialogs

• US 20030130841A1
• Filed: 12/05/2002
• Published: 07/10/2003
• Est. Priority Date: 12/07/2001
• Status: Active Grant

• Current Assignee: Microsoft Technology Licensing LLC (Microsoft Corporation)
• Original Assignee: AT&T Corporation (AT&T, Inc.)
• Inventors: Gupta, Narendra K., Bangalore, Srinivas, Rahim, Mazin G.
• Granted Patent: US 8,190,436 B2
• Time in Patent Office: Days
• US Class Current: 704/231
• CPC Class Codes: G06F 40/30 Semantic analysis
• CPC Class Codes: G10L 13/00 Speech synthesis; Text to speech systems
• CPC Class Codes: G10L 15/02 Feature extraction for speech recognition; Selection of recognition unit
• CPC Class Codes: G10L 15/1815 Semantic context, e.g. disambiguation of the recognition hypotheses based on word meaning
• CPC Class Codes: G10L 15/1822 Parsing for meaning understanding
• CPC Class Codes: G10L 15/26 Speech to text systems G10L15/08 takes precedence

Abstract

A system and method are disclosed that improve automatic speech recognition in a spoken dialog system. The method comprises partitioning speech recognizer output into self-contained clauses, identifying a dialog act in each of the self-contained clauses, qualifying dialog acts by identifying a current domain object and/or a current domain action, and determining whether further qualification is possible for the current domain object and/or current domain action. If further qualification is possible, then the method comprises identifying another domain action and/or another domain object associated with the current domain object and/or current domain action, reassigning the another domain action and/or another domain object as the current domain action and/or current domain object and then recursively qualifying the new current domain action and/or current object. This process continues until nothing is left to qualify.

Claims

1. A method of understanding a spoken utterance in a spoken dialog system, the method comprising:

partitioning a recognizer output associated with a spoken utterance into self-contained clauses;

identifying a dialog act for each of the self-contained clauses; and

qualifying each dialog act to generate a semantic representation of the spoken utterance for use in dialog management.

2. The method of understanding a spoken utterance of claim 1, wherein the step of qualifying each dialog act further comprises:

identifying a domain-dependent object and/or a domain-dependent action.

3. The method of understanding a spoken utterance of claim 2, wherein the step of qualifying each dialog act further comprises:

recursively qualifying each identified domain-dependent object and domain-dependent action until nothing is left to qualify.

4. The method of understanding a spoken utterance of claim 1, wherein the dialog act identified for each of the self-contained clauses is a domain-independent dialog act.

5. A method of understanding a spoken utterance in a spoken dialog system, the method comprising:

applying a first domain-independent module to partition a speech recognizer output associated with a spoken utterance into self-contained clauses;

applying a second domain-independent module to identify dialog acts within the self-contained clauses; and

applying a domain-dependent module to qualify the identified dialog acts.

6. The method of claim 5, wherein qualifying the identified dialog act further comprises identifying a domain-dependent action and/or a domain-dependent object associated with the self-contained clauses.

7. The method of claim 6, further comprising:

recursively qualifying each identified domain-dependent action and/or domain-dependent object until nothing is left to qualify.

8. A method of processing data in a spoken dialog system, the method comprising:

(1) partitioning speech recognizer output into self-contained clauses;

(2) identifying a dialog act in each of the self-contained clauses;

(3) qualifying the dialog act by identifying a current domain object and/or a current domain action;

(4) determining whether further qualification is possible for the current domain object and/or current domain action; and

(5) if further qualification is possible;

(a) identifying another domain action and/or another domain object associated with the current domain object and/or current domain action;

(b) reassigning the another domain action and/or another domain object as the current domain action and/or current domain object; and

(c) returning to step (3).

9. A module used in a spoken dialog system, the module receiving data from a speech recognizer, the module comprising:

a clausifier that identifies clauses within received speech recognition data;

a dialog act identifier that associates a dialog act to each identified clause;

a dialog act qualifier that identifies domain-dependent actions and/or domain-dependent objects in each identified clause, wherein the module outputs a semantic representation of the received speech recognition data.

10. The module of claim 9, wherein the dialog act associated with each identified clause is a domain-independent dialog act.

11. The module of claim 10, wherein the module further performs a recursive process wherein, when possible, each identified domain-dependent action and/or domain-dependent object is further qualified to identify another domain-dependent action and/or domain-dependent object.

12. The module of claim 9, wherein the semantic representation of the received speech recognition data is used by a dialog manager in a spoken dialog system to determine a system response to a use input.

13. A computer-readable medium storing instructions related to a spoken dialog system to control a computer device to perform the steps of:

partitioning a recognizer output associated with a spoken utterance into self-contained clauses;

identifying a dialog act for each of the self-contained clauses; and

qualifying each dialog act to generate a semantic representation of the spoken utterance for use in dialog management.

14. The computer-readable medium of claim 13, wherein the step of qualifying each dialog act further comprises:

identifying a domain-dependent object and/or a domain-dependent action.

15. The computer-readable medium of claim 14, wherein the step of qualifying each dialog act further comprises:

recursively qualifying each identified domain-dependent object and domain-dependent action until nothing is left to qualify.

16. The computer-readable medium of claim 13, wherein the dialog act identified for each of the self-contained clauses is a domain-independent dialog act.

17. A computer-readable medium storing instructions related to a spoken dialog system to control a computer device to perform the steps of:

applying a first domain-independent module to partition a speech recognizer output associated with a spoken utterance into self-contained clauses;

applying a second domain-independent module to identify dialog acts within the self-contained clauses; and

applying a domain-dependent module to qualify the identified dialog acts.

18. The computer-readable medium of claim 17, wherein qualifying the identified dialog acts further comprises identifying a domain-dependent action and/or a domain-dependent object associated with the self-contained clauses.

19. The computer-readable medium of claim 18, wherein the instructions further comprise the step of:

recursively qualifying each identified domain-dependent action and/or domain-dependent object until nothing is left to qualify.

20. A computer-readable medium storing instructions related to a spoken dialog system to control a computer device to perform the steps of:

(1) partitioning speech recognizer output into self-contained clauses;

(2) identifying a dialog act in each of the self-contained clauses;

(3) qualifying the dialog act by identifying a current domain-dependent object and/or a current domain-dependent action;

(4) determining whether further qualification is possible for the current domain-dependent object and/or current domain-dependent action; and

(5) if further qualification is possible;

(a) identifying another domain-dependent action and/or another domain-dependent object associated with the current domain object and/or current domain-dependent action;

(b) reassigning the another domain-dependent action and/or another domain-dependent object as the current domain action and/or current domain-dependent object; and

(c) returning to step (3).