Knowledge management using text classification

• US 7,673,234 B2
• Filed: 03/11/2002
• Issued: 03/02/2010
• Est. Priority Date: 03/11/2002
• Status: Active Grant

• Current Assignee: The Boeing Co.
• Original Assignee: The Boeing Co.
• Inventors: Woods, Steven D., Quach, Lesley Hong-Minh, Kao, Anne S. W., Poteet, Stephen R.
• Primary Examiner(s): RUTLEDGE, AMELIA L
• Application Number:
  US10/095,673
• Publication Number:
  US 20030172357A1
• Time in Patent Office: 2,913 Days
• Field of Search: 715/529, 715/255, 715/256
• US Class Current: 715/255
• CPC Class Codes: G06F 16/353 into predefined classes

Abstract

A method and system are provided for determining metadata to be assigned to a document. A document is identified to be published to a collection of documents. A category for publishing the document to the collection is automatically recommended using a text classifier. At least one category for the document may be selected, and the selected category is assigned as metadata for the document. A method and system are also provided for constructing a search for a document in a collection of documents. A search query is entered for a document in a collection of documents. A search category for searching for the document in the collection of documents is automatically recommended using a text classifier. At least one search category for searching for the document is selected.

Claims

1. A computer executable method for determining metadata to be assigned to a document, the method comprising:

identifying a document to be published to a collection of documents organized according to a predetermined taxonomy sanctioned by an organization independent of an entity that controls the collection of documents and that includes a stored plurality of predetermined knowledge categories;

applying a trained text classifier to the identified document;

outputting from the trained text classifier at least one of the plurality of predetermined knowledge categories;

without parsing the document into predicate structures, automatically presenting to a user the at least one of the plurality of predetermined knowledge categories that is output by the trained text classifier as at least one recommended knowledge category for the user to select to publish the document to the collection of documents;

receiving, from the user, a selection of at least one knowledge category for the document;

expanding a knowledge category to at least one related knowledge category, wherein the at least one related knowledge category is within a hierarchy in the predetermined taxonomy of categories including the at least one recommended knowledge category;

receiving, from the user, a selection of at least one related knowledge category for the document; and

assigning the at least one selected knowledge category and the at least one related knowledge category as metadata for publishing the document to the collection of documents.

2. The method of claim 1, wherein automatically presenting at least one recommended knowledge category further includes:

reading text of the document;

retrieving the stored plurality of predetermined knowledge categories;

associating the text of the document with the plurality of predetermined knowledge categories; and

recommending at least one of the predetermined plurality of knowledge categories as the at least one recommended knowledge category when, association of the text of the document with the at least one recommended knowledge category exceeds a predetermined threshold.

3. The method of claim 2, wherein the text of the document is associated with the at least one recommended knowledge category when a numerical value for the association exceeds a predetermined probability that the at least one recommended knowledge category applies to the document.

4. The method of claim 2, wherein the text of the document is associated with the at least one recommended knowledge category when the text of the document includes the at least one recommended knowledge category.

5. The method of claim 1, wherein the trained text classifier is a Naive Bayes text classifier.

6. The method of claim 1, wherein the at least one knowledge category selected by the user for the document includes the at least one recommended knowledge category recommended by the trained text classifier.

7. The method of claim 1, wherein the at least one knowledge category selected by the user for the document does not include the at least one recommended knowledge category recommended by the trained text classifier.

8. The method of claim 1, wherein the predetermined knowledge categories include at least one controlled list of keywords selected by an industry specific group.

9. The method of claim 8, wherein the at least one controlled list of keywords includes keywords selected by a professional organization.

10. The method of claim 1, further comprising expanding a particular knowledge category of the at least one recommended knowledge category output by the trained text classifier to at least one stochastically related knowledge category, wherein the at least one stochastically related knowledge category is not related to the particular knowledge category by the predetermined taxonomy but has previously been recommended with the particular knowledge category by the trained text classifier.

11. A system for determining metadata to be assigned to a document, the system comprising:

at least one input interface configured to input a document to be published to a collection of documents organized according to a predetermined taxonomy that includes a stored plurality of predetermined knowledge categories, wherein the plurality of predetermined knowledge categories includes at least one controlled list of keywords;

a user interface configured to permit a user to select at least one knowledge category for the document as a selected knowledge category; and

a processor including;

a first component configured to apply a trained text classifier to the inputted document and to output from the trained text classifier at least one of the plurality of predetermined knowledge categories as a recommended knowledge category;

a second component configured to automatically present to a user the recommended knowledge category and at least one related knowledge category including at least one of a specialized related knowledge category and a generalized related knowledge category within a hierarchy of categories within the predetermined taxonomy including the at least one recommended knowledge category, wherein the user is permitted to select at least one of the recommended knowledge category and the at least one related knowledge category; and

a third component configured to assign the at least one selected knowledge category as metadata for publishing the document to the collection of documents.

12. The system of claim 11, wherein the second component further includes:

a fourth component configured to read text of the document;

a fifth component configured to retrieve the stored plurality of predetermined knowledge categories;

a sixth component configured to associate the text of the document with the plurality of predetermined knowledge categories; and

a seventh component configured to recommend at least one of the predetermined plurality of knowledge categories as the at least one recommended knowledge category when association of the text of the document with the at least one recommended knowledge category exceeds a predetermined threshold.

13. The system of claim 12, wherein the sixth component includes a Naive Bayes text classifier.

14. The system of claim 12, wherein the text of the document is associated with the at least one, recommended knowledge category when a numerical value for the association exceeds a predetermined probability that the at least one recommended knowledge category applies to the document.

15. The system of claim 12, wherein the text of the document is associated with the at least one recommended knowledge category when the text of the document includes the at least one recommended knowledge category.

16. The system of claim 11, wherein the at least one knowledge category selected by the user for the document includes the at least one recommended knowledge category recommended by the trained text classifier.

17. The system of claim 11, wherein the at least one knowledge category selected by the user for the document does not include the at least one recommended knowledge category recommended by the trained text classifier.

18. The system of claim 11, wherein the processor further includes an eighth component configured to expand a knowledge category to at least one related knowledge category.

19. The system of claim 11, wherein the at least one controlled list of keywords includes terms selected by an industry organization.

20. The system of claim 11, wherein the at least one controlled list of keywords includes keywords selected by a professional organization.

21. The system of claim 11, wherein the at least one related knowledge category includes at least one stochastically related knowledge category, wherein the at least one stochastically related knowledge category is not related to the knowledge category by the predetermined taxonomy but has previously been recommended with the knowledge category by the trained text classifier.

22. A computer executable method for determining metadata to be assigned to a document, the method comprising:

identifying a document to be published to a collection of documents organized according to a hierarchical taxonomy, wherein each of a plurality of documents included in the collection of documents is associated with one or more knowledge categories included in the hierarchical taxonomy;

applying a trained text classifier to the identified document, wherein the trained text classifier identifies one or more recommended knowledge categories from the hierarchical taxonomy based on content of the identified document;

presenting the one or more recommended knowledge categories to a user;

enabling the user to expand at least one knowledge category included in the one or more recommended knowledge categories to present one or more hierarchically related knowledge categories included in the hierarchical taxonomy;

enabling the user to make a selection of at least one selected knowledge category for the document from the one or more recommended knowledge categories and the one or more hierarchically related knowledge categories; and

assigning the at least one selected knowledge category as metadata in publishing the document to the collection of documents.

23. The computer executable method of claim 22, further comprising automatically assigning the one or more recommended knowledge categories as the metadata in publishing the document to the collection of documents when the user fails to provide the selection of the at least one selected knowledge category.

24. The computer executable method of claim 23, wherein the one or more hierarchically related knowledge categories includes one or more generalized related knowledge categories at a next higher level in the hierarchical taxonomy than the at least one knowledge category.

25. The computer executable method of claim 23, wherein the one or more hierarchically related knowledge categories include one or more specialized related knowledge categories within the at least one knowledge category.

26. The computer executable method of claim 25, further comprising presenting one or more related knowledge categories to the user, wherein the one or more related knowledge categories are related to one of the one or more specialized related knowledge categories within the at least one knowledge category and wherein the one or more related knowledge categories are selectable as the at least one selected knowledge category.

27. The computer executable method of claim 22, further comprising enabling the at least one selected knowledge category to be removed from association with the document.

28. The computer executable method of claim 22, further comprising receiving, from the user, a search string of the hierarchical taxonomy for a non-recommended knowledge category.

29. The computer executable method of claim 28, further comprising receiving, from the user, a selection of the non-recommended knowledge category as the at least one selected knowledge category.

30. The computer executable method of claim 22, wherein the one or more recommended knowledge categories are presented in a first terms box and the at least one selected knowledge category is presented in a second terms box.

31. A computer executable method for determining metadata to be assigned to a document, the method comprising:

identifying a document to be published to a collection of documents organized according to a hierarchical taxonomy, wherein each of a plurality of documents included in the collection of documents is associated with one or more knowledge categories included in the hierarchical taxonomy;

applying a trained text classifier to the identified document, wherein the trained text classifier identifies one or more recommended knowledge categories from the hierarchical taxonomy based on content of the identified document;

presenting the one or more recommended knowledge categories to a user;

enabling the user to expand at least one knowledge category included in the one or more recommended knowledge categories to present one or more related knowledge categories included in the hierarchical taxonomy, wherein the related knowledge categories include;

one or more knowledge categories at a next more general level in the hierarchical taxonomy than the at least one knowledge category; and

one or more knowledge categories at a next more specific level in the hierarchical taxonomy than the at least one knowledge category;

enabling the user to make a selection of at least one selected knowledge category for the document from the one or more recommended knowledge categories and the one or more related knowledge categories; and

assigning the at least one selected knowledge category as metadata in publishing the document to the collection of documents.

32. The computer executable method of claim 31, further comprising enabling the user to expand the at least one knowledge category included in the recommended knowledge categories to present one or more non-hierarchically related knowledge categories.

33. The computer executable method of claim 32, wherein a relationship between the knowledge category and the one or more non-hierarchically related knowledge categories includes a rule-based relationship.

34. The computer executable method of claim 33, wherein the rule-based relationship relates the at least one knowledge category the one or more non-hierarchically related knowledge categories because the at least one knowledge category and the one or more non-hierarchically related knowledge categories share a common author.

35. The computer executable method of claim 33, wherein the rule-based relationship relates the at least one knowledge category and the one or more non-hierarchically related knowledge categories because the at least one knowledge category and the one or more non-hierarchically related knowledge categories share a common technology.

36. The computer executable method of claim 33, wherein the rule-based relationship relates the at least one knowledge category and the one or more non-hierarchically related knowledge categories because the at least one knowledge category and the one or more non-hierarchically related knowledge categories share a common application.

37. The computer executable method of claim 33, wherein the rule-based relationship relates the at least one knowledge category and the one or more non-hierarchically related knowledge categories because the at least one knowledge category and the one or more non-hierarchically related knowledge categories share a common domain.

38. The computer executable method of claim 31, further comprising automatically assigning the one or more recommended knowledge categories as the metadata in publishing the document to the collection of documents when the user fails to provide the selection of the at least one selected knowledge category.

39. The computer executable method of claim 31, further comprising enabling the at least one selected knowledge category to be removed from association with the document.

40. The computer executable method of claim 31, further comprising receiving, from the user, a search string of the hierarchical taxonomy for a non-recommended knowledge category.

41. The computer executable method of claim 40, further comprising receiving, from the user, a selection of the non-recommended knowledge category as the at least one selected knowledge category.

42. The computer executable method of claim 31, wherein the one or more recommended knowledge categories are presented in a first terms box and the at least one selected knowledge category is presented in a second terms box.