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

• US 7,627,096 B2
• Filed: 01/14/2005
• Issued: 12/01/2009
• Est. Priority Date: 01/14/2005
• Status: Active Grant

• Current Assignee: Microsoft Technology Licensing LLC (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:
  US11/036,201
• Publication Number:
  US 20060161431A1
• Time in Patent Office: 1,782 Days
• Field of Search: 704235-236, 704 1- 7, 704/231, 704246-247, 704251-257, 704270-271, 704275-276, 704E15001-E15007, 704/E15.014, 704E15019-E15024, 370/351, 370/355, 379 70- 72, 379/80, 379 8801- 8828, 379 9001- 9323
• US Class Current: 379/88.03
• 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 method for processing a call is disclosed. The method receives a speech input via a call and transforms at the speech input into a textual format. The method also creates a list of salient terms of actions and objects from the text, adjusts the confidence level of objects on the list if the dominant term is an action and selects a complimentary object from the list to combine with the action to form an action-object pair. The method further adjusts a confidence level of actions on the list if the dominant term is an object and selects a complementary action from the list to combine with the action to form the action-object pair, and routes the call based on the action-object pair.

Claims

1. A method of routing a call based on speech input via the call, the method comprising:

transforming the speech input via the call into text;

creating a list comprising terms including actions and objects that are included based at least partially on the text;

identifying a dominant term in the list;

when the dominant term is an action, determining a corresponding confidence level of each object in the list, wherein the corresponding confidence level of the action is based at least in part on the dominant term;

when the dominant term is an object, determining the corresponding confidence level of each action in the list, wherein the corresponding confidence level of the object is based at least in part on the dominant term;

forming an action-object pair by selecting a particular term from the list to be a complementary term and pairing the complementary term with the dominant term; and

routing the call based on the action-object pair.

2. The method of claim 1, further comprising, when the dominant term is an action, masking actions within the list that are not the dominant term to prevent the masked actions from being selected to be the complementary term, and masking objects within the list that are identified as invalid objects to prevent any masked object from being selected to be the complementary term.

3. The method of claim 1, when the dominant term is an object, further comprising masking objects that are not the dominant term within the list to prevent the masked objects from being selected to be the complementary term, and masking actions within the list that are identified as invalid actions to prevent any masked action from being selected to be the complementary term.

4. The method of claim 1, further comprising assigning a corresponding confidence level to each action in the list and to each object in the list.

5. The method of claim 4, wherein when the dominant term is an action, determining the corresponding confidence level of each object in the list further comprises adjusting the corresponding confidence level of a first object in the list based at least in part on a likelihood that the first object is consistent with the dominant term.

6. The method of claim 4, wherein when the dominant term is an object, determining the corresponding confidence level of each action in the list further comprises adjusting the corresponding confidence level of a first action in the list based at least in part on a likelihood that the first action is consistent with the dominant term.

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

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

9. The method of claim 1, further comprising:

mapping a phrase within the text to a corresponding action within a synonym table; and

including the corresponding action in the list.

10. The method of claim 1, further comprising:

mapping a phrase within the text to a corresponding object within a synonym table; and

including the corresponding object in the list.

11. The method of claim 1, wherein the associated confidence level of a first term is adjusted utilizing a synonym table.

12. The system of claim 1, wherein selecting a particular term from the list to be the complementary term comprises:

when the dominant term is an action, selecting a particular object based at least in part on the corresponding confidence level of the particular object; and

when the dominant term is an object, selecting a particular action based at least in part on the corresponding confidence level of the particular action.

13. A system of routing a call based upon speech input, the system comprising:

a processor;

a call routing module to;

transform speech received into text;

create a list comprising terms including actions and objects based at least partially on the text;

identify a term in the list to be a dominant term that is one of an object and an action;

determine an associated confidence level of each object in the list when the dominant term is an action; and

determine the associated confidence level of each action in the list when the dominant term is an object;

wherein each associated confidence level is based at least in part on the dominant term;

a pairing module associated with the call routing system and operable to create an action-object pair by selecting a particular term from the list to be a complementary term and pairing the complementary term with the dominant term, wherein;

when the dominant term is an action the particular term selected is one of the objects;

when the dominant term is an object the particular term selected is one of the actions; and

the particular term is selected based at least in part on the associated confidence level of the particular term;

and

a switch to route the call based on the action-object pair.

14. The system of claim 13, wherein the call routing module is to assign the associated confidence level having a corresponding initial value to each of the terms in the list, and wherein each associated confidence level represents a corresponding probability that the corresponding term represents an intent of the caller.

15. The system of claim 14, wherein the call routing module is configured to adjust the associated confidence level of at least one of the terms in the list based upon the dominant term identified.

16. The system of claim 13, wherein the dominant term is selected based at least in part on the associated confidence level of each action and of each object in the list.

17. The system of claim 16, wherein the associated confidence level of the dominant term has a greater value than the associated confidence level of other terms in the list.

18. A system to determine a routing destination based upon speech, the system comprising:

an acoustic engine configured to accept a speech input and to output a text based on at least a portion of the speech;

a semantic engine coupled to the acoustic engine and operable to identify one or more actions and one or more objects based at least in part on the text;

a probability system operable to assign a corresponding confidence level to each of the one or more actions and the one or more objects identified;

an action-object pairing system to;

identify a dominant entry from the one or more identified actions and the one or more identified objects; and

select a complement to the dominant entry from the one or more identified actions and the one or more identified objects, wherein the selection is based at least in part on the dominant entry; and

form an action-object pair that includes the dominant entry and the complement to the dominant entry;

and

an action-object routing table operable to provide a routing destination based at least partially on the action-object pair.

19. The system of claim 18, wherein the action is selected from a list of expected utterances comprising acquire, cancel, change, inquire, inform, and how to use.

20. The system of claim 18, wherein the object is selected from a list of expected utterances comprising DSL, basic service, call notes, caller ID, bill payment, other providers, coupon specials, names and number, and store locations.

21. The system of claim 18, further comprising memory storing a synonym table including a plurality of expected actions and expected objects.

22. The system of claim 18, wherein the complement to the dominant entry is selected based at least in part on a likelihood of consistency with the dominant entry.

23. The system of claim 18, wherein the action-object pairing system is to mask a particular term in a list of the one or more actions and the one or more objects identified prior to selecting the complement to the dominant entry, wherein masking comprises preventing the particular term from being selected to be the complement to the dominant entry.

24. The system of claim 23, wherein when the dominant term is one of the objects in the list the particular term that is masked is another of the objects in the list, and when the dominant term is one of the actions in the list the particular term that is masked is another of the actions in the list.

25. The system of claim 23, wherein the particular term that is masked is inconsistent with the dominant term.