System and method for independently recognizing and selecting actions and objects in a speech recognition system

• US 7,966,176 B2
• Filed: 10/22/2009
• Issued: 06/21/2011
• Est. Priority Date: 01/14/2005
• Status: Expired due to Fees

• Current Assignee: Nuance Communications, Inc. (Microsoft Corporation)
• Original Assignee: AT&T Intellectual Property I LP (AT&T, Inc.)
• Inventors: Bushey, Robert R., Sabourin, Michael, Potvin, Carl, Knott, Benjamin Anthony, Martin, John Mills
• Primary Examiner(s): Wozniak, James S
• Assistant Examiner(s): KOVACEK, DAVID M
• Application Number:
  US12/603,619
• Publication Number:
  US 20100040207A1
• Time in Patent Office: 607 Days
• Field of Search: 704 1- 7, 704 46-247, 704251-257, 704270-271, 704275-276, 704E15001-E15007, 704/E15.014, 704E15019-E15024, 370/351, 370/355, 397 70- 72, 397/80, 397 8801- 8828, 397 9001- 9323
• US Class Current: 704/231
• CPC Class Codes: G10L 15/08 Speech classification or search
• CPC Class Codes: G10L 15/1822 Parsing for meaning understanding
• CPC Class Codes: G10L 2015/0631 Creating reference templates; Clustering

Abstract

A system includes an acoustic input engine configured to accept a speech input, to recognize phonemes of the speech input, and to create word strings based on the recognized phonemes. The system includes a semantic engine coupled to the acoustic engine and operable to identify actions and to identify objects by parsing the word strings. The system also includes an action-object pairing system to identify a dominant entry from the identified actions and the identified objects, to select a complement to the dominant entry from the identified actions and the identified objects, and to form an action-object pair that includes the dominant entry and the complement. The system further includes an action-object routing table operable to provide a routing destination based on the action-object pair. The system also includes a call routing module to route a call to the routing destination.

Claims

1. A computer-implemented method, comprising:

recognizing, at a call routing system comprising a processor and a memory, phonemes of an utterance using an acoustical model;

creating word strings based on the recognized phonemes of the utterance;

parsing the word strings to identify actions and objects;

creating a table comprising terms, the terms including the actions and the objects;

identifying a dominant term of the table via the processor;

when the dominant term is an action, masking actions of the table that are not the dominant term and masking objects of the table that are identified as invalid objects with reference to the dominant term;

when the dominant term is an object, masking objects of the table that are not the dominant term, masking actions of the table that are identified as invalid actions with reference to the dominant term, determining a corresponding confidence level of each action of the table based on the dominant term, and modifying the corresponding confidence level of each action of the table based on a likelihood that the action is consistent with the dominant term;

selecting a particular term from the table as a complementary term and pairing the complementary term with the dominant term to form an action-object pair; and

routing a call received by the call routing system based on the action-object pair.

2. The computer-implemented method of claim 1, further comprising, when the dominant term is an action, determining a corresponding confidence level of each object of the table, wherein the corresponding confidence level of each object is based on the dominant term.

3. The computer-implemented method of claim 2, wherein, when the dominant term is an action, determining the corresponding confidence level of each object of the table further comprises modifying the corresponding confidence level of each object of the table based on a likelihood that the object is consistent with the dominant term.

4. The computer-implemented method of claim 1, further comprising assigning the corresponding confidence level to each action of the table.

5. The computer-implemented method of claim 1, wherein the actions include at least one of a verb and an adverb-verb combination.

6. The computer-implemented method of claim 1, wherein the objects include at least one of a noun and an adjective-noun combination.

7. The computer-implemented method of claim 1, wherein each of the confidence levels represents a probability that a corresponding term represents an intent of a caller.

8. The computer-implemented method of claim 1, further comprising adjusting the confidence level associated with at least one of the terms in the table based upon the dominant term.

9. A system, comprising:

a processor;

a memory including a call routing module that, when executed by the processor, causes the processor to;

receive a speech utterance;

identify a term of a table as a dominant term, the table comprising terms including actions and objects based on the speech utterance;

when the dominant term is an action, mask actions of the table that are not the dominant term and mask objects of the table that are invalid objects with reference to the dominant term; and

when the dominant term is an object, mask objects of the table that are not the dominant term, mask actions of the table that are invalid actions with reference to the dominant term, determine a confidence level of each action of the table based on the dominant term, and modify the confidence level of each action of the table based on a likelihood that the action is consistent with the dominant term; and

a pairing module to select a particular term from the table as a complementary term and pair the complementary term with the dominant term to form an action-object pair,

wherein the call routing module is executable by the processor to route a call to a destination based on the action-object pair.

10. The system of claim 9, wherein the call routing module is operable to assign the confidence level to each object of the table when the dominant term is an action and to assign the confidence level to each action of the table when the dominant term is an object, wherein each confidence level is based at least in part on the dominant term.

11. The system of claim 10, wherein each of the confidence levels represents a probability that a corresponding term represents an intent of a caller.

12. The system of claim 10, wherein the dominant term is selected based at least in part on the confidence level of each action and of each object in the table.

13. The system of claim 9, wherein the call routing module is configured to adjust the associated confidence level of at least one of the terms in the table based upon the dominant term identified.

14. A system, comprising:

an acoustic input engine configured to accept a speech input, to recognize phonemes of the speech input, and to create word strings based on the recognized phonemes;

a semantic engine coupled to the acoustic engine and operable to identify actions and to identify objects by parsing the word strings;

an action-object pairing system to identify a dominant entry from the identified actions and the identified objects, to select a complement to the dominant entry, to form an action-object pair that includes the dominant entry and the complement, and when the dominant entry is an object, to mask objects of the table that are not the dominant entry, to mask actions of the table that are invalid actions with reference to the dominant entry, to determine a confidence level of each action of the table based on the dominant entry, and to modify the confidence level of each action of the table based on a likelihood that the action is consistent with the dominant entry;

an action-object routing table operable to provide a routing destination based on the action-object pair; and

a call routing module to route a call to the routing destination.

15. The system of claim 14, wherein the identified actions include at least one of acquire, cancel, change, inquire, inform, and how to use.

16. The system of claim 14, wherein the identified objects include at least one of digital subscriber line (DSL), basic service, call notes, caller identification (ID), bill payment, other providers, coupon specials, names and number, and store locations.

17. The system of claim 14, wherein the complement to the dominant entry is selected based on a likelihood of consistency with the dominant entry.

18. The system of claim 14, wherein, before selecting the complement to the dominant entry, the action-object pairing system is operable to mask at least one of the identified actions and at least one of the identified objects.

19. The system of claim 18, wherein the identified actions that are not the dominant entry are masked and wherein the identified objects that are determined to be invalid objects with reference to the dominant entry are masked when the dominant entry is an action.

20. The system of claim 18, wherein the identified objects that are not the dominant entry are masked and wherein the identified actions that are determined to be invalid actions with reference to the dominant entry are masked when the dominant entry is an object.