DETERMINATION OF A BASIS FOR A NEW DOMAIN MODEL BASED ON A PLURALITY OF LEARNED MODELS

• US 20130018825A1
• Filed: 07/11/2011
• Published: 01/17/2013
• Est. Priority Date: 07/11/2011
• Status: Active Grant

• Current Assignee: Accenture Global Services Limited (Accenture PLC)
• Original Assignee: Accenture Global Services Limited (Accenture PLC)
• Inventors: GHANI, Rayid, Krema, Marko
• Granted Patent: US 8,620,837 B2
• Time in Patent Office: Days
• US Class Current: 706/12
• CPC Class Codes: G06N 20/00 Machine learning

Abstract

In a machine learning system in which a plurality of learned models, each corresponding to a unique domain, already exist, new domain input for training a new domain model may be provided. Statistical characteristics of features in the new domain input are first determined. The resulting new domain statistical characteristics are then compared with statistical characteristics of features in prior input previously provided for training at least some of the plurality of learned models. Thereafter, at least one learned model of the plurality of learned models is identified as the basis for the new domain model when the new domain input statistical characteristics compare favorably with the statistical characteristics of the features in the prior input corresponding to the at least one learned model.

Claims

1. A method, in a machine learning system comprising a plurality of learned models uniquely corresponding to a plurality of domains, for determining which of the plurality of learned models to use as a basis for a new domain model, the method comprising:

determining, by a processing device, statistical characteristics of features in new domain input to provide new feature statistical characteristics, wherein the new domain input is provided for training the new domain model;

comparing, by the processing device, the new feature statistical characteristics with statistical characteristics of at least some of the plurality of learned models; and

identifying, by the processing device, at least one learned model of the plurality of learned models as the basis for the new domain model when the new feature statistical characteristics compare favorably with the statistical characteristics of the features in the prior input corresponding to the at least one learned model.

2. The method of claim 1, wherein the statistical characteristics of the features in the new domain input and the statistical characteristics of the features in the prior input are based on frequency of occurrence of the features in the new domain input and in the prior input.

3. The method of claim 1, wherein the statistical characteristics of the at least some of the plurality of learned models are based on prior input for training the at least some of the plurality of learned models.

4. The method of claim 1, wherein comparing the new feature statistical characteristics with the statistical characteristics of the features in the prior input further comprises comparing based on a Hellinger metric.

5. The method of claim 1, wherein a first learned model of the plurality of learned models comprises a combination of other learned models of the plurality of models.

6. The method of claim 1, wherein a first learned model of the plurality of learned models comprises a subset of another learned model of the plurality of models.

7. The method of claim 1, wherein the new domain input comprises text.

8. The method of claim 1, further comprising:

providing, by the processing device, the new domain model based on the at least one learned model and the new domain input.

9. The method of claim 8, further comprising providing the new domain model based on active learning input.

10. The method of claim 8, further comprising:

analyzing, by the processing device, additional new domain input based on the new domain model.

11. The method of claim 10, wherein the additional new domain input comprises text.

12. An apparatus, for use in a machine learning system comprising a plurality of learned models uniquely corresponding to a plurality of domains, for determining which of the plurality of learned models to use as a basis for a new domain model, the apparatus comprising:

a processor;

a storage device, operatively connected to the processor, and having stored thereon instructions that, when executed by the processor, cause the processor to;

determine statistical characteristics of features in new domain input to provide new feature statistical characteristics, wherein the new domain input is provided for training the new domain model;

compare the new feature statistical characteristics with statistical characteristics of at least some of the plurality of learned models; and

identify at least one learned model of the plurality of learned models as the basis for the new domain model when the new feature statistical characteristics compare favorably with the statistical characteristics of the features in the prior input corresponding to the at least one learned model.

13. The apparatus of claim 12, wherein the statistical characteristics of the features in the new domain input and the statistical characteristics of the features in the prior input are based on frequency of occurrence of the features in the new domain input and in the prior input.

14. The apparatus of claim 12, wherein the statistical characteristics of the at least some of the plurality of learned models are based on prior input for training the at least some of the plurality of learned models.

15. The apparatus of claim 12, wherein those instructions that, when executed, cause the processor to compare the new feature statistical characteristics with the statistical characteristics of the features in the prior input are further operative to compare based on a Hellinger metric.

16. The apparatus of claim 12, wherein a first learned model of the plurality of learned models comprises a combination of other learned models of the plurality of models.

17. The apparatus of claim 12, wherein a first learned model of the plurality of learned models comprises a subset of another learned model of the plurality of models.

18. The apparatus of claim 12, wherein the new domain input comprises text.

19. The apparatus of claim 12, the storage device further comprising instructions that, when executed by the processor, cause the processor to:

provide the new domain model based on the at least one learned model and the new domain input.

20. The apparatus of claim 19, wherein those instructions that, when executed by the processor, cause the processor to provide the new domain model are further operative to provide the new domain model based on active learning input.

21. The apparatus of claim 20, the storage device further comprising instructions that, when executed by the processor, cause the processor to:

analyze additional new domain input based on the new domain model.

22. The apparatus of claim 21, wherein the additional new domain input comprises text.

23. An apparatus, for use in a machine learning system comprising a plurality of learned models uniquely corresponding to a plurality of domains, for determining which of the plurality of learned models to use as a basis for a new domain model, the apparatus comprising:

a statistical determination component operative to determine statistical characteristics of features in prior input provided for training at least some of the plurality of learned models and to determine statistical characteristics of features in new domain input to provide new feature statistical characteristics, wherein the new domain input is provided for training the new domain model;

a comparator, operatively connected to the statistical determination component, that compares the new feature statistical characteristics with the statistical characteristics of the features in the prior input and identifies at least one learned model of the plurality of learned models as the basis for the new domain model when the new feature statistical characteristics compare favorably with the statistical characteristics of the features in the prior input corresponding to the at least one learned model.

24. The apparatus of claim 23, wherein the statistical determination component is further operative, when determining the statistical characteristics of features in the new domain input and the statistical features in the prior input, to determine frequency of occurrence of the features in the new domain input and the prior input.

25. The apparatus of claim 23, wherein the comparator compares the new feature statistical characteristics with the statistical characteristics of the features in the prior input based on a Hellinger metric.

26. The apparatus of claim 23, wherein the new domain input and the prior input comprise text.

27. The apparatus of claim 23, further comprising:

a training component, operatively connected to the comparator, operative to provide the new domain model based on the at least one learned model and the new domain input.

28. The apparatus of claim 27, further comprising:

an analysis component, configured to receive the new domain model, operative to analyze additional new domain input based on the new domain model.

29. The apparatus of claim 28, wherein the additional new domain input comprises text.