Method and apparatus for training a text classifier

• US 5,675,710 A
• Filed: 06/07/1995
• Issued: 10/07/1997
• Est. Priority Date: 06/07/1995
• Status: Expired due to Term

• Current Assignee: Alcatel-Lucent USA, Inc. (Nokia Corporation)
• Original Assignee: Lucent Technologies, Inc. (Nokia Corporation)
• Inventors: Lewis, David Dolan
• Primary Examiner(s): Downs, Robert W.
• Application Number:
  US08/484,436
• Time in Patent Office: 853 Days
• Field of Search: 395/10, 395/600, 395/613, 364/419.19, 364/419.07, 382/159, 382/229
• US Class Current: 706/12
• CPC Class Codes: G06F 16/353 into predefined classes
• CPC Class Codes: G06F 18/00 Pattern recognition
• CPC Class Codes: Y10S 706/934 Information retrieval or Information management

Abstract

A method and apparatus for training a text classifier is disclosed. A supervised learning system and an annotation system are operated cooperatively to produce a classification vector which can be used to classify documents with respect to a defined class. The annotation system automatically annotates documents with a degree of relevance annotation to produce machine annotated data. The degree of relevance annotation represents the degree to which the document belongs to the defined class. This machine annotated data is used as input to the supervised learning system. In addition to the machine annotated data, the supervised learning system can also receive manually annotated data and/or a user request. The machine annotated data, along with the manually annotated data and/or the user request, are used by the supervised learning system to produce a classification vector. In one embodiment, the supervised learning system comprises a relevance feedback mechanism. The relevance feedback mechanism is operated cooperatively with the annotation system for multiple iterations until a classification vector of acceptable accuracy is produced. The classification vector produced by the invention is the result of a combination of supervised and unsupervised learning.

Claims

1. A method for training a classifier to classify at least one document which has not been manually annotated with respect to a defined class;

performing an operation on data including a retrieval status value associated with the document to generate at least one parameter value;

calculating a degree of relevance representing the degree to which said document belongs to said defined class, said degree of relevance being a function of at least the retrieval status value and the parameter value; and

training said classifier using said degree of relevance.

2. The method of claim 1 further comprising the steps of:

automatically annotating said at least one document with said calculated degree of relevance to produce at least one automatically annotated document; and

wherein said step of training said classifier further comprises the step of training said classifier using said at least one automatically annotated document.

3. The method of claim 2 wherein said step of training said classifier using said at least one automatically annotated document further comprises the step of:

performing a relevance feedback function using said at least one automatically annotated document to produce a classification vector.

4. The method of claim 1 wherein said defined class is defined by a request.

5. The method of claim 1 wherein said defined class is defined by at least one manually annotated document.

6. The method of claim 1 wherein said defined class is defined by both a request and at least one manually annotated document.

7. The method of claim 1 wherein said degree of relevance is proportional to an estimate of the probability of said at least one document in said first set being relevant to said defined class.

8. The method of claim 1 wherein the operation includes logistic regression.

9. The method of claim 1 wherein the data includes logistic training data.

10. The method of claim 1 wherein the parameter value includes a value of a logistic parameter.

11. A method for producing a classification vector for use in classifying at least one non-manually annotated document with respect to a defined class, said method comprising the steps of:

performing an operation on data including a retrieval status value associated with the document to generate at least one parameter value;

calculating a degree of relevance representing the degree to which said non-manually annotated document belongs to the defined class, said degree of relevance being a function of at least the retrieval status value and the parameter value;

automatically annotating said non-manually annotated document with said degree of relevance to produce a machine annotated document; and

performing a relevance feedback function using said machine annotated document.

12. The method of claim 11 wherein said step of calculating a degree of relevance further comprises the step of:

calculating an estimate of the probability that said non-manually annotated document belongs to the defined class.

13. The method of claim 11 wherein said step of performing a relevance feedback function further comprises the step of:

performing the relevance feedback function using a manually annotated document.

14. The method of claim 11 wherein the operation includes logistic regression.

15. The method of claim 11 wherein the data includes logistic training data.

16. The method of claim 11 wherein the parameter value includes a value of a logistic parameter.

29. The apparatus of claim 15 wherein the parameter value includes a value of a logistic parameter.

17. An apparatus for training a classifier to classify at least one document which has not been manually annotated with respect to a defined class, said apparatus comprising:

an operating processor for performing an operation on data including a retrieval status value associated with the document to generate at least one parameter value;

an annotation processor for automatically annotating the document to produce at least one automatically annotated document, said annotation including a degree of relevance representing the degree to which said at least one automatically annotated document belongs to said defined class, said degree of relevance being a function of at least the retrieval status value and the parameter value; and

a supervised learning processor for training the classifier using said at least one automatically annotated document.

18. The system of claim 17 wherein said defined class is defined by a request.

19. The system of claim 17 wherein said supervised learning system further comprises:

a relevance feedback module.

20. The system of claim 17 wherein said annotation system further comprises:

classification means for calculating a retrieval status value for said at least one document; and

means responsive to said classification means for calculating said degree of relevance for said at least one document.

21. The apparatus of claim 17 wherein the operation includes logistic regression.

22. The apparatus of claim 17 wherein the data includes logistic training data.

23. The apparatus of claim 17 wherein the parameter value includes a value of a logistic parameter.

24. An apparatus for training a classifier to classify at least one document which has not been manually annotated with respect to a defined class, said apparatus comprising:

means for performing an operation on data including a retrieval status value associated with the document to generate at least one parameter value;

means for calculating a degree of relevance representing the degree to which the document belongs to said defined class, said degree of relevance being a function of at least the retrieval status value and the parameter value; and

means for training the classifier using said degree of relevance.

25. The apparatus of claim 24 further comprising:

annotation means for automatically annotating said at least one document with said calculated degree of relevance to produce at least one automatically annotated document; and

wherein said means for training the classifier comprises a relevance feedback mechanism responsive to said annotation means for producing a classification vector.

26. The apparatus of claim 24 wherein said degree of relevance is proportional to an estimate of the probability of said at least one document being relevant to said defined class.

27. The apparatus of claim 24 wherein the operation includes logistic regression.

28. The apparatus of claim 24 wherein the data includes logistic training data.