Learning system for classification of terminology

• US 6,101,515 A
• Filed: 05/31/1996
• Issued: 08/08/2000
• Est. Priority Date: 05/31/1996
• Status: Expired due to Term

• Current Assignee: Oracle International Corporation (Oracle Corporation)
• Original Assignee: Oracle Corporation
• Inventors: Wical, Kelly, Conklin, James
• Primary Examiner(s): Au, Amelia
• Assistant Examiner(s): Wu, Jingge
• Application Number:
  US08/655,957
• Time in Patent Office: 1,530 Days
• Field of Search: 707/500, 707/530, 707/531, 707/532, 707/204, 707/1, 707/2, 707/3, 707/4, 707/5, 707/101, 707/907, 706/12, 706/46, 706/934, 382/159, 382/155, 382/224, 382/225, 382/226, 382/227
• US Class Current: 715/234
• CPC Class Codes: G06F 16/353 into predefined classes

Abstract

A learning system learns terms in the context of a set of documents. During an accumulation phase, the learning system accumulates contextual data for a term in the form of a categorization schema. The categorization schema, which is based on a classification hierarchy, classifies the term in categories such that the classifications are based on uses of the terms in the set of documents During a computational phase, the learning system analyzes the categorization schema, and selects, if sufficient contextual data has been accumulated, a single category in the classification system to classify the term. A content processing system, which understands the thematic content of documents, is used in conjunction with the learning system.

Claims

1. A method for automated learning of terminology from an input set of documents, the method comprising the steps of:

storing a classification system comprising a plurality of categories of terminology arranged to reflect associations among related categories;

processing the input set of documents to classify each of the documents into a category of the classification system, wherein the documents classified contain a term for learning;

generating contextual data for the term for learning by mapping the categories selected to the classification system so as reflect associations among the categories selected;

analyzing the contextual data, including analyzing all of the categories selected in the classification system for the term, prior to selecting a single category for the term; and

selecting a single category, based on the associations of the categories in the classification system, for the term to learn the term as the single category.

2. The method as set forth in claim 1, wherein the step of analyzing the contextual data comprises the steps of:

determining whether the term is ambiguous such that the contextual data for the term does not indicate selection of a single category; and

removing the term from consideration for learning if the term is ambiguous.

3. A method for automated learning of at least one term from an input set of documents, the method comprising the steps of:

storing a classification system comprising a plurality of categories of terminology arranged in a hierarchy of categories so as to reflect associations among related categories;

processing the input set of documents to generate contextual data by classifying the term in a plurality of categories of the classification system; and

analyzing the contextual data by performing hierarchical clustering analysis on the selected categories of the hierarchy of categories in the classification system to identify a cluster of categories and to select a single category in the cluster from the plurality of selected categories to learn the term as the single category selected.

4. The method as set forth in claim 3, wherein:

the step of processing the input set of documents to generate contextual data comprises the step of generating a categorization schema from said classification system comprising a plurality of nodes, arranged in a plurality of levels of the hierarchy and logically coupled by at least one branch, and data associated with each node, wherein the nodes identify the categories classified; and

the step of analyzing the contextual data comprises the step of analyzing the categorization schema, including the data, to select a node in the categorization schema to classify the term.

5. The method as set forth in claim 4, wherein the step of analyzing the categorization schema comprises the steps of:

analyzing nodes, including the data, from the top of at least one branch of the categorization schema to nodes further down on the branch so as to determine classification of the term in a node on the branch; and

selecting a node as far down on the branch in the categorization schema to classify the term as supported by the data.

6. The method as set forth in claim 4, wherein the step of analyzing the categorization schema comprises the steps of:

accumulating the data at each node of the categorization schema;

generating a weighted average at each node of the categorization schema based on the data; and

utilizing the weighted average at each node to select a node in the categorization schema to classify the term.

7. The method as set forth in claim 4, wherein the step of analyzing the categorization schema comprises the steps of:

selecting a branch in the categorization schema that includes a cluster of nodes; and

analyzing the branch, starting from the top of the branch, to select a node supported by the data.

8. The method as set forth in claim 4, wherein the step of analyzing the categorization schema comprises the step of eliminating from consideration nodes in the categorization schema that include insufficient data to support classification of a term into a node.

9. The method as set forth in claim 4, wherein the data associated with each node specifies a number of documents classified to each node and a classification strength associating the node with the term.

10. The method as set forth in claim 9, wherein the step of analyzing the categorization schema comprises the step of analyzing the classification strength associating the term with the node being analyzed.

11. The method as set forth in claim 3, wherein the step of storing a classification system comprises the step of storing the classification system such that associated categories are distanced to reflect a semantic and lexical relationship between respective categories.

12. The method as set forth in claim 4, wherein the step of analyzing the categorization schema comprises the steps of analyzing the nodes of the categorization schema to identify clusters and eliminating nodes from consideration if the nodes are located outside the cluster.

13. The method as set forth in claim 3, wherein the step of analyzing the contextual data comprises the steps of:

determining whether the term is ambiguous such that the contextual data for the term does not indicate classification in a single category; and

removing the term from consideration for learning if the term is ambiguous.

14. The method as set forth in claim 3, further comprising the step of removing inappropriate terms from consideration from learning.

15. A computer readable medium having a set of instructions stored therein, which when executed by a computer, causes the computer to perform the steps of:

storing a classification system comprising a plurality of categories of terminology arranged in a hierarchy of categories so as to reflect associations among related categories;

processing the input set of documents to generate contextual data by classifying the term in a plurality of categories of the classification system; and

analyzing the contextual data by performing hierarchical clustering analysis on the selected categories of the hierarchy of categories in the classification system to identify a cluster of categories and to select a single category in the cluster from the plurality of selected categories to learn the term as the single category selected.

16. The computer readable medium as set forth in claim 15, wherein:

the instructions for processing the input set of documents to generate contextual data comprises instructions for generating a categorization schema from said classification system comprising a plurality of nodes, arranged in a plurality of levels of the hierarchy and logically coupled by at least one branch, and data associated with each node, wherein the nodes identify the categories classified; and

the instructions for analyzing the contextual data comprises instructions for analyzing the categorization schema, including the data, to select a node in the categorization schema to classify the term.

17. The computer readable medium as set forth in claim 16, wherein the instructions for analyzing the categorization schema comprises instructions for:

analyzing nodes, including the data, from the top of at least one branch of the categorization schema to nodes further down on the branch so as to determine classification of the term in a node on the branch; and

selecting a node as far down on the branch in the categorization schema to classify the term as supported by the data.

18. The computer readable medium as set forth in claim 16, wherein instructions for analyzing the categorization schema comprises instructions for:

accumulating the data at each node of the categorization schema;

generating a weighted average at each node of the categorization schema based on the data; and

utilizing the weighted average at each node to select a node in the categorization schema to classify the term.

19. The computer readable medium as set forth in claim 16, wherein instructions for analyzing the categorization schema comprises instructions for:

selecting a branch in the categorization schema that includes a cluster of nodes; and

analyzing the branch, starting from the top of the branch, to select a node supported by the data.

20. The computer readable medium as set forth in claim 16, wherein the instructions for analyzing the categorization schema comprises instructions for eliminating from consideration nodes in the categorization schema that include insufficient data to support classification of a term into a node.

21. The computer readable medium as set forth in claim 16, wherein the data associated with each node specifies a number of documents classified to each node and a classification strength associating the node with the term.

22. The computer readable medium as set forth in claim 21, wherein the instructions for analyzing the categorization schema comprises instructions for analyzing the classification strength associating the term with the node being analyzed.

23. The computer readable medium as set forth in claim 16, wherein instructions for storing a classification system comprises instructions for storing the classification system such that associated categories are distanced to reflect a semantic and lexical relationship between respective categories.

24. The computer readable medium as set forth in claim 16, wherein the instructions for analyzing the categorization schema comprises instructions for analyzing the nodes of the categorization schema to identify clusters and eliminating nodes from consideration if the nodes are located outside the cluster.

25. The computer readable medium as set forth in claim 15, wherein the instructions for analyzing the contextual data comprises instructions for:

determining whether the term is ambiguous such that the contextual data for the term does not indicate classification in a single category; and

removing the term from consideration for learning if the term is ambiguous.

26. The computer readable medium as set forth in claim 15, further comprising instructions for removing inappropriate terms from consideration from learning.

27. A text processing system for learning at least one term in an input set of documents comprising:

a classification system comprising a plurality of categories of terminology arranged to reflect associations among the categories;

a content processing system for processing the input set of documents to generate a thematic profile for each document that identifies a category for classification of the documents that include the term for learning; and

a learning system coupled to receive the thematic profile and categories of the classification system for generating a categorization schema by classifying the term into a plurality of different categories in the classification system, and for analyzing the categorization schema to select a single category from the plurality of different categories in the classification system to learn the term as the single category selected.

28. A computer readable medium having a set of instructions stored therein, which when executed by a computer, causes the computer to perform the steps of:

storing a classification system comprising a plurality of categories of terminology arranged to reflect associations among the categories;

processing the input set of documents to select a plurality of categories for a plurality of documents, one category for each document, to classify the documents, wherein the documents classified include a term for learning;

generating contextual data for the term for learning by mapping the categories selected to the classification system so as reflect associations among said category selected;

analyzing the contextual data, including analyzing all of the categories selected in the classification system for the term, prior to selecting a single category for the term; and

selecting a single category from the plurality of categories in the contextual data for the term to learn the term as the single category.