Disambiguation of themes in a document classification system

• US 5,930,788 A
• Filed: 07/17/1997
• Issued: 07/27/1999
• Est. Priority Date: 07/17/1997
• Status: Expired due to Term

• Current Assignee: Oracle International Corporation (Oracle Corporation)
• Original Assignee: Oracle Corporation
• Inventors: Wical, Kelly
• Primary Examiner(s): Black, Thomas G.
• Assistant Examiner(s): Mizrahi, Diane D.
• Application Number:
  US08/895,773
• Time in Patent Office: 740 Days
• Field of Search: 707/4, 707/5, 707/6, 707/7, 707/104, 707/500, 707/513, 707/530, 382/112, 382/175, 382/178, 382/190, 382/202, 380/21, 395/109, 395/117, 395/200.34
• US Class Current: 1/1
• CPC Class Codes: G06F 16/353 into predefined classes
• CPC Class Codes: Y10S 707/99934 Query formulation, input preparation, or translation
• CPC Class Codes: Y10S 707/99935 Query augmenting and refining, e.g. inexact access
• CPC Class Codes: Y10S 707/99936 Pattern matching access
• CPC Class Codes: Y10S 707/99937 Sorting
• CPC Class Codes: Y10S 707/99945 Object-oriented database structure processing

Abstract

A document classification system includes disambiguation processing to validate categories that have been preliminarily classified for themes of a document. The themes of a document are preliminarily classified through use of a classification hierarchy that contains a plurality of categories. The disambiguation processing determines, for a theme selected for disambiguation, whether the category preliminarily classified for the theme selected is valid by analyzing the relationships among the category preliminarily classified for the theme and other categories classified for different themes in the document. The disambiguation processing also utilizes a category cross reference database, which comprises a list of category cross reference pairs, to disambiguate categories assigned to themes by pairing a category classified for a theme and other categories classified for other themes in the document and by comparing these category pairs with category cross reference database pairs. If a match occurs, then the categories of a document category pair are validated.

Claims

1. A computer-implemented method for validating a category classified for a theme in a document classification system, said method comprising the steps of:

processing at least one document by generating a thematic context output that indicates applicability of a plurality of thematic constructions, and by generating a plurality of themes from said thematic context output, wherein said themes define the overall content of said document;

storing a classification hierarchy that includes a plurality of categories;

receiving a plurality of categories that preliminarily classify said themes for said document;

selecting a theme from a document to validate a category preliminarily classified for said theme; and

determining whether said category preliminarily classified for said theme selected is valid by analyzing relationships among said category preliminarily classified for said theme and other categories classified for different themes in said document.

2. The method as set forth in claim 1, wherein:

the step of storing a classification hierarchy comprisies the step of storing a classification hierarchy with said categories arranged in a hierarchical relationship; and

the step of determining whether said preliminary classification for said theme is valid comprises the steps of;

identifying a category preliminarily classified for a first theme of said document that is a child category to a parent category preliminarily classified for a second theme of said document;

generating a theme strength for each theme, wherein said theme strength reflects a relative contribution of the overall content for said document;

comparing a theme strength of said first theme corresponding to said child category with a theme strength of said second theme corresponding to said parent category; and

validating said child category as a classification for said first theme and validating said parent category as a classification for said second theme if said theme strength for said second theme is greater than said theme strength for said first theme.

3. The method as set forth in claim 1, wherein the step of determining whether said preliminary classification for said theme is valid comprises the steps of:

determining whether a theme identified for said document is also a category preliminarily classified for a different theme of said document;

determining whether a theme identified for said document is also a parent category of a category classified for a different theme of said document; and

validating a category classified for a theme if said category, or a parent category for said category, is classified for a different theme.

4. The method as set forth in claim 1, wherein the step of determining whether said preliminary classification for said theme is valid comprises the steps of:

generating a theme strength for each theme, wherein said theme strength reflects a relative contribution of the overall content for said document;

generating a total theme strength for a category by summing all theme strengths for all themes classified in said category;

generating a total theme strength for a parent category of said category by summing all theme strengths for all themes classified under said parent category;

comparing said total theme strength for said parent category with said total theme strength for said category; and

validating said category as a classification for said themes if said total theme strength for said parent category is greater than said total theme strength for said parent category.

5. A computer-implemented method for validating a category preliminarily classified for theme in a document classification system, said method comprising the steps of:

processing at least one document by generating a thematic context output that indicates applicability of a plurality of thematic constructions, and by generating a plurality of themes from said thematic context output, wherein said themes define the overall content of said document;

receiving a preliminary classification of categories for said themes in said document;

storing a category cross reference database that comprises a list of category cross reference pairs, wherein categories of a category pair have a semantic, linguistic, or use association;

selecting a theme from said document to validate a category preliminarily classified for said theme;

generating document category pairs by combining a category classified for said theme and each category classified for other themes in said document;

comparing said document category pairs with category cross reference database pairs to determine if a match occurs; and

validating categories of a document category pair if a match between said document category pair and said category cross reference database pairs occurs.

6. The method as set forth in claim 5, further comprising the steps of:

extracting, from a classification hierarchy, parent categories for each category preliminarily classified for said themes;

generating parent document category pairs by combining a parent category for said category classified for said theme and each parent category for other categories classified for other themes;

comparing said parent document category pairs with category cross reference database pairs to determine if a match occurs; and

validating a preliminary classification of said theme in a category if a match occurs.

7. The method as set forth in claim 5, further comprising the steps of:

extracting, from a classification hierarchy, parent categories for each category classified for said themes;

generating parent/child document category pairs by combining a parent category for said category classified for said theme and each category classified for other themes and by combining said category classified for said theme and each parent category for other categories classified for other themes;

comparing said parent/child document category pairs with cross reference database category pairs to determine if a match occurs; and

validating classification of said theme in said category if a match occurs.

8. A computer-implemented method for classifying themes in a document, said method comprising the steps of:

determining whether themes in said document are completely ambiguous;

assigning a category for each theme to preliminarily classify each theme under said category if a theme is not completely ambiguous; and

determining, for themes assigned a category, whether a category classified for a theme is valid by analyzing other categories classified for other themes of said document.

9. A computer readable medium comprising a plurality of instructions, which when executed by a computer, causes the computer to perform the steps of:

processing at least one document by generating a thematic context output that indicates applicability of a plurality of thematic constructions, and by generating a plurality of themes from said thematic context output, wherein said themes define the overall content of said document;

storing a classification hierarchy that includes a plurality of categories;

receiving a plurality of categories that preliminarily classify said themes for said document;

selecting a theme from a document to validate a category preliminarily classified for said theme; and

determining whether said category preliminarily classified for said theme selected is valid by analyzing relationships among said category preliminarily classified for said theme and other categories classified for different themes in said document.

10. The computer readable medium as set forth in claim 9, wherein:

the instructions for storing a classification hierarchy comprise instructions for storing a classification hierarchy with said categories arranged in a hierarchical relationship; and

the instructions for determining whether said preliminary classification for said theme is valid comprise instructions for;

identifying a category preliminarily classified for a first theme of said document that is a child category to a parent category preliminarily classified for a second theme of said document;

generating a theme strength for each theme, wherein said theme strength reflects a relative contribution of the overall content for said document;

comparing a theme strength of said first theme corresponding to said child category with a theme strength of said second theme corresponding to said parent category; and

validating said child category as a classification for said first theme and validating said parent category as a classification for said second theme if said theme strength for said second theme is greater than said theme strength for said first theme.

11. The computer readable medium as set forth in claim 9, wherein the instructions for determining whether said preliminary classification for said theme is valid comprise instructions for:

determining whether a theme identified for said document is also a category preliminarily classified for a different theme of said document;

determining whether a theme identified for said document is also a parent category of a category classified for a different theme of said document; and

validating a category classified for a theme if said category, or a parent category for said category, is classified for a different theme.

12. The computer readable medium as set forth in claim 9, wherein the instructions for determining whether said preliminary classification for said theme is valid comprise instructions for:

generating a theme strength for each theme, wherein said theme strength reflects a relative contribution of the overall content for said document;

generating a total theme strength for a category by summing all theme strengths for all themes classified in said category;

generating a total theme strength for a parent category of said category by summing all theme strengths for all themes classified under said parent category;

comparing said total theme strength for said parent category with said total theme strength for said category; and

validating said category as a classification for said themes if said total theme strength for said parent category is greater than said total theme strength for said parent category.

13. A computer readable medium comprising a plurality of instructions, which when executed by a computer, causes the computer to perform the steps of:

processing at least one document by generating a thematic context output that indicates applicability of a plurality of thematic constructions, and by generating a plurality of themes from said thematic context output, wherein said themes define the overall content of said document;

receiving a preliminary classification of categories for said themes in said document;

storing a category cross reference database that comprises a list of category cross reference pairs, wherein categories of a category pair have a semantic, linguistic, or use association;

selecting a theme from said document to validate a category preliminarily classified for said theme;

generating document category pairs by combining a category classified for said theme and each category classified for other themes in said document;

comparing said document category pairs with category cross reference database pairs to determine if a match occurs; and

validating categories of a document category pair if a match between said document category pair and said category cross reference database pairs occurs.

14. The computer readable medium as set forth in claim 13, further comprising instructions for:

extracting, from a classification hierarchy, parent categories for each category preliminarily classified for said themes;

generating parent document category pairs by combining a parent category for said category classified for said theme and each parent category for other categories classified for other themes;

comparing said parent document category pairs with category cross reference database pairs to determine if a match occurs; and

validating a preliminary classification of said theme in a category if a match occurs.

15. The computer readable medium as set forth in claim 13, further comprising instructions for:

extracting, from a classification hierarchy, parent categories for each category classified for said themes;

generating parent/child document category pairs by combining a parent category for said category classified for said theme and each category classified for other themes and by combining said category classified for said theme and each parent category for other categories classified for other themes;

comparing said parent/child document category pairs with cross reference database category pairs to determine if a match occurs; and

validating classification of said theme in said category if a match occurs.

16. A computer readable medium comprising a plurality of instructions, which when executed by a computer, causes the computer to perform the steps of:

receiving a document comprising a plurality of themes;

determining whether said themes in a document are completely ambiguous;

assigning a category for each theme to preliminarily classify each theme under said category if a theme is not completely ambiguous; and

determining, for themes assigned a category, whether a category classified for a theme is valid by analyzing other categories classified for other themes of said document.

17. A computer system comprising:

memory for storing a plurality of themes for a document;

processor unit for determining whether said themes in said document are completely ambiguous, for assigning a category for each theme to preliminarily classify each theme under said category if a theme is not completely ambiguous, and for determining, for themes assigned a category, whether a category classified for a theme is valid by analyzing other categories classified for other themes of said document.