Robust personalization through biased regularization

• US 7,886,266 B2
• Filed: 04/06/2006
• Issued: 02/08/2011
• Est. Priority Date: 04/06/2006
• Status: Active Grant

• Current Assignee: Microsoft Technology Licensing LLC (Microsoft Corporation)
• Original Assignee: Microsoft Corporation
• Inventors: Kienzle, Wolf, Chellapilla, Kumar H.
• Primary Examiner(s): Opsasnick; Michael N
• Application Number:
  US11/278,949
• Publication Number:
  US 20070239450A1
• Time in Patent Office: 1,769 Days
• Field of Search: 707/7, 717/104
• US Class Current: 717/104
• CPC Class Codes: G10L 15/07 to the speaker

Abstract

The subject disclosure pertains to systems and methods for personalization of a recognizer. In general, recognizers can be used to classify input data. During personalization, a recognizer is provided with samples specific to a user, entity or format to improve performance for the specific user, entity or format. Biased regularization can be utilized during personalization to maintain recognizer performance for non-user specific input. In one aspect, regularization can be biased to the original parameters of the recognizer, such that the recognizer is not modified excessively during personalization.

Claims

1. A system for enhancing performance of a data pattern recognizer for an entity, the system comprising:

one or more processors;

a memory;

a base parameter component that obtains a base set of parameters of the recognizer;

a personalization component that updates the recognizer based at least in part upon at least one sample customized to the entity, the personalization component utilizes regularization favoring the base set of parameters of the recognizer during update of the recognizer;

a regularization component that utilizes regularization to determine a new set of parameters for the recognizer; and

a parameter update component that updates the base set of parameters of the recognizer to the new set of parameters,

wherein one or more of the components comprise a set of processor-executable instructions stored in the memory and that, when executed by the one or more processors, perform its described operation.

2. The system of claim 1, the regularization component comprises:

a deviation component that determines the deviation from the base set of parameters to a possible set of parameters;

an error rate component that determines the error rate over the at least one customized sample using the possible set of parameters; and

a minimization component that selects the new set of parameters to minimize the sum of the error rate and the deviation.

3. The system of claim 1, the base parameter component maintains the base set of parameters of the recognizer.

4. The system of claim 1, the base parameter component obtains the base set of parameters of the recognizer from the recognizer.

5. The system of claim 1, the recognizer classifies image data, the image data includes at least one of printed text and handwritten text.

6. The system of claim 1, further comprising:

a sample receiving component that receives the at least one customized sample, the at least one customized sample includes raw data and a corresponding label.

7. The system of claim 6, the personalization component adds the corresponding label of the customized sample to a set of recognized labels of the recognizer.

8. The system of claim 1, the regularization utilized is Tichonov regularization.

9. A method for enhancing the performance of a trained recognizer for an individual, the method comprising:

retraining, by one or more computing devices configured to enhance the performance of the trained recognizer for the individual, the trained recognizer based at least in part upon at least one sample personalized to the individual using regularization biased to a base state of the trained recognizer;

determining a new state of the trained recognizer based upon the regularization and the at least one personalized sample; and

updating the trained recognizer, the determining of the new state of the trained recognizer further comprises;

determining a distance between the base state of the trained recognizer and a possible state of the trained recognizer;

determining risk for the at least one personalized sample using the possible state of the trained recognizer; and

minimizing the distance and the risk.

10. The method of claim 9, the at least one personalized sample includes image data, the image data includes at least one of printed text and handwritten text.

11. The method of claim 9, retraining the trained recognizer is based at least in part on a weight for the at least one personalized sample, the weight based in part on the likelihood of recognition error for the personalized sample.

12. The method of claim 9, further comprising maintaining the base state of the trained recognizer.

13. The method of claim 9, further comprising obtaining the base state of the trained recognizer from the trained recognizer.

14. A system for enhancing performance of a data-pattern recognizer for an individual, the system comprising:

one or more processors;

a memory;

a personalization component that customizes the recognizer to the individual, the personalization being based at least in part upon a sample generated by the individual, the personalization component utilizes regularization biased to a base set of parameters of the recognizer;

a base-parameter component that obtains the base set of parameters of the recognizer; and

a regularization component that utilizes regularization to determine a new set of parameters for the recognizer based upon the regularization and the sample,

wherein one or more of the components comprise a set of processor-executable instructions stored in the memory that, when executed by the one or more processors, perform its described operation.

15. The system of claim 14, further comprising a parameter-update component that updates the base set of parameters of the recognizer to the new set of parameters.

16. The system of claim 15, further comprising a base-parameter component that obtains the base set of parameters of the recognizer for the personalization component to customize the recognizer to the individual and maintains the base set of parameters once the base set of parameters are updated by the parameter-update component.

17. The system of claim 14, further comprising:

a deviation component that determines the deviation from the base set of parameters to a possible set of parameters;

an error-rate component that determines the error rate over the sample using the possible set of parameters; and

a minimization component that selects the new set of parameters to minimize the sum of the error rate and the deviation.

18. The system of claim 14, wherein the base set of parameters include a base set of weights of the recognizer and the new set of parameters include a new set of weights of the recognizer.

19. The system of claim 14, the recognizer classifies image data or audio data.