Recognition results postprocessor for use in voice recognition systems

• US 7,899,671 B2
• Filed: 02/05/2004
• Issued: 03/01/2011
• Est. Priority Date: 02/05/2004
• Status: Active Grant

• Current Assignee: Arlington Technologies, LLC (Dominion Harbor Enterprises, LLC)
• Original Assignee: Avaya Incorporated
• Inventors: Sanders, Derek, Tokarev, Vladimir Sergeyevich, Cooper, Robert S.
• Primary Examiner(s): Wozniak; James S
• Assistant Examiner(s): KOVACEK, DAVID M
• Application Number:
  US10/772,488
• Publication Number:
  US 20050177376A1
• Time in Patent Office: 2,581 Days
• Field of Search: 704231-233, 704243-257, 706/12, 706 14- 16, 706/21, 706 26- 28, 379 671- 72, 379 74- 77, 379 79- 82, 379 8801- 8806, 379 8819- 8828
• US Class Current: 704/252
• CPC Class Codes: G10L 15/065 Adaptation

Abstract

Systems and techniques for analyzing voice recognition results in order to improve efficiency and accuracy of voice recognition. When a voice activated module undertakes a voice recognition attempt, it invokes a voice recognition module that constructs a list of voice recognition results. The list is analyzed by a results postprocessor that employs information relating to past recognition results and user information to make changes to the list. The results postprocessor may delete results that have been previously rejected during a current recognition transaction and may further alter and reorder the results list based on historical results. The results postprocessor may further alter and reorder the results list based on information relating to the user engaging in the recognition attempt.

Claims

1. A voice recognition system comprising:

a plurality of voice activated modules for receiving voice recognition results representing voice inputs from a user and taking actions in response to the voice inputs;

a voice recognition module for receiving voice inputs from a user and performing voice recognition on the voice inputs, performing voice recognition on a voice input comprising identifying members of a collection of elements representing potential matches to the voice input, the voice recognition module being operative to prepare a list of potential voice recognition results for a voice input under consideration, each of the potential voice recognition results representing a candidate for a result matching the voice input received from the user and having an associated confidence level;

a results postprocessor for processing the list of potential voice recognition results to improve speed and accuracy of voice recognition, the results postprocessor being operative to make changes to the list based on information relating to past results of recognition attempts in order to associate a higher priority with members of the list having a higher likelihood of matching the voice input under consideration as indicated by the past results of recognition attempts; and

wherein the processing of the list of potential voice recognition results comprises comparing the associated confidence levels of the potential voice recognition results to an upper threshold and a lower threshold, wherein the potential voice recognition results are kept if the associated confidence levels are above the upper threshold, the potential voice recognition results are not kept if the associated confidence levels are below the lower threshold, and potential voice recognition results with associated confidence levels between the upper threshold and the lower threshold are only kept if they match any of the past results of recognition attempts.

2. The system of claim 1, wherein the results postprocessor is further operative to make changes to the list based on previously stored information relating to expected user actions.

3. The system of claim 2, wherein the previously stored information relating to expected user actions comprise a user model, the user model comprising a likely contact cache including entries for contacts the user is estimated to be likely to call.

4. The system of claim 3, wherein making changes to the list includes reordering the list if an entry in the result list corresponds to an entry in the likely contact cache.

5. The system of claim 4, wherein the information relating to past results of recognition attempts includes information relating to results of a current recognition transaction.

6. The system of claim 5, wherein the information relating to past results of recognition attempts includes a skip list indicating rejected results during the current recognition transaction, and wherein making changes to the list includes removing results appearing in the skip list from the result list.

7. The system of claim 6, wherein the results postprocessor is operative to receive parameters from a module engaging in a voice recognition attempt and to adapt its operation based on the parameters.

8. A results postprocessor for improving efficiency and accuracy of voice recognition, comprising:

a repository of information relating to a current voice recognition attempt; and

a processing module for processing potential voice recognition results in a result list, the potential voice recognition results representing members of a collection of elements representing potential matches to a voice input under consideration in a current voice recognition attempt, each potential voice recognition result having an associated confidence level, the processing module being operative to examine information in the repository relating to the current voice recognition attempt and to make changes to a results list compiled in response to a voice input, the changes being made based on the information stored in the repository, the changes associating a higher priority with results in the results list that are indicated to have a higher priority of matching the voice input based on information in the repository relating the characteristics of elements of data collection; and

wherein the processing of the list of potential voice recognition results comprises comparing the associated confidence levels of the potential voice recognition results to an upper threshold and a lower threshold, wherein the potential voice recognition results are kept if the associated confidence levels are above the upper threshold, the potential voice recognition results are not kept if the associated confidence levels are below the lower threshold, and the potential voice recognition results with associated confidence levels between the upper threshold and the lower threshold are only kept if they match any of the past results of recognition attempts.

9. The results postprocessor of claim 8, wherein the changes to the results list include removing results that have been rejected during a current recognition transaction.

10. The results postprocessor of claim 9, wherein the processing module is further operative to retrieve user and historical information and to make changes to the results list based on the user and historical information.

11. The results postprocessor of claim 10, wherein the user information includes a likely contact cache including contacts the user is estimated as likely to call and the changes to the result list include reordering the results list if an entry in the results list corresponds to an entry in the likely contact cache.

12. A method of analyzing voice recognition results, comprising the steps of:

examining a list of voice recognition results representing candidates for matches to a voice input from a user, the candidates being members of a collection of data selected as potential matches to a particular voice input under consideration, each voice recognition result associated with a confidence level, the list being complied in response to the voice input under consideration;

making changes to the list of voice recognition results using a processor based on information relating to results of past recognition attempts, to associate a higher selection priority with members of the list indicated as having a higher priority of matching the voice input based on the results of past recognition attempts;

wherein the making changes to the list of voice recognition results comprises comparing associated confidence levels of the voice recognition results to an upper threshold and a lower threshold, wherein the voice recognition results are kept if the associated confidence levels are above the upper threshold, the voice recognition results are not kept if the associated confidence levels are below the lower threshold, and the voice recognition results with an associated confidence level between the upper threshold and the lower threshold are only kept if they match any of the past results of recognition attempts.

13. The method of claim 12, wherein the step of making changes to the list based on information relating to results of past recognition attempts includes removing results previously rejected during a current voice recognition transaction.

14. The method of claim 13, wherein the step of making changes to the list based on information relating to results of past recognition attempts includes reordering the list based on historical recognition results.

15. The method of claim 14, wherein the step of making changes to the list based on information relating to results of past recognition attempts includes making changes to the list based on information relating to a user engaging in a current voice recognition transaction.

16. The method of claim 15, wherein the step of making changes to the list based on information relating to a user engaging in a current voice recognition transaction includes examining a likely contact cache indicating contact the user is estimated as likely to call and reordering the list if an entry in the list corresponds to an entry in the likely contact cache.