Method and apparatus for ontology-based classification of media content

• US 7,873,642 B2
• Filed: 01/09/2008
• Issued: 01/18/2011
• Est. Priority Date: 08/03/2004
• Status: Active Grant

• Current Assignee: SINOEAST CONCEPT LIMITED (Tencent Holdings Limited)
• Original Assignee: International Business Machines Corporation
• Inventors: Smith, John R., Tseng, Belle L., Wu, Yi
• Primary Examiner(s): Fleurantin; Jean B.
• Assistant Examiner(s): GOFMAN, ALEX N
• Application Number:
  US11/971,252
• Publication Number:
  US 20080133466A1
• Time in Patent Office: 1,105 Days
• Field of Search: 707/688, 707/748, 707/999.107
• US Class Current: 707/748
• CPC Class Codes: G06F 18/254 of classification results, e.g. of results related to same input data
• CPC Class Codes: G06V 10/809 of classification results, e.g. where the classifiers operate on the same input data
• CPC Class Codes: G06V 20/40 in video content extracting overlay text G06V20/62; video retrieval G06F16/70; processing of video elementary streams in video servers H04N21/234; processing of video elementary streams in video clients H04N21/44

Abstract

A method and apparatus for ontology-based classification of media content are provided. With the method and apparatus, initial confidence values of classifiers in a hierarchical classification structure are modified based on relationships between classifiers. A confidence value for a classifier is boosted by a boosting factor based on a correspondence between the confidence value and confidence values of ancestor classifiers in the hierarchical classification structure. A confidence value for a classifier is modified by a confusion factor based on a correspondence between the confidence value of the classifier and the confidence values of mutually exclusive classifiers in the hierarchical classification structure. In this way, a more accurate representation of the actual confidence that media content falls within the classification associated with the classifier is obtained. From this improved classification mechanism, indices for media content may be generated for use in accessing the media content at a later time.

Claims

1. A method, in a data processing system, for classifying media data, comprising:

responsive to receiving input media data, classifying the input media data for later access based on classification;

applying a plurality of classifiers in a hierarchical classification structure to the input media data;

generating confidence scores for the plurality of classifiers in the hierarchical classification structure;

modifying a confidence score automatically for a classifier in the hierarchical classification structure based on a relationship between the classifier and other classifiers in the hierarchical classification structure and the confidence scores of the other classifiers in the hierarchical classification structure by applying a confusion factor to the confidence score based on a correspondence between the confidence score and confidence scores of mutually exclusive classifiers in the hierarchical classification structure;

generating a representation of the input media data based on the modified confidence score being higher than a predetermined threshold confidence score; and

storing the generated representation of the input media data.

2. The method of claim 1, wherein modifying the confidence score automatically for a classifier includes boosting the confidence score using a boosting factor determined based on a correspondence between the confidence score and confidence scores of ancestor classifiers in the hierarchical classification structure.

3. The method of claim 2, wherein the boosting factor is determined as a function of a difference between the confidence score of the classifier and confidence scores of each ancestor classifier in the hierarchical classification structure.

4. The method of claim 1, wherein generating a representation of the input media data includes generating an index data structure based on the modified confidence score.

5. The method of claim 1, further comprising:

using the stored representation of the input media data to access the input media data.

6. The method of claim 5, wherein using the stored representation of the input media data to access the input media data includes:

receiving a search query;

determining if terms in the search query match terms in the stored representation of the input media data; and

accessing the input media data if the terms in the search query match terms in the stored representation of the input media data.

7. The method of claim 1, wherein the confusion factor is defined as a reciprocal of a difference between the confidence score of the classifier and a maximum confidence score from mutually exclusive classifiers in the hierarchical classification structure.

8. The method of claim 1, wherein the confusion factor is a probability of misclassifying the input media data into a classifier in the plurality of classifiers.

9. A computer program product in a recordable-type computer readable medium for classifying media data, comprising:

first instructions for classifying input media data for later access based on classification in response to receiving the input media data;

second instructions for applying a plurality of classifiers in a hierarchical classification structure to the input media data;

third instructions for generating confidence scores for the plurality of classifiers in the hierarchical classification structure;

fourth instructions for modifying a confidence score automatically for a classifier in the hierarchical classification structure based on a relationship between the classifier and other classifiers in the hierarchical classification structure and the confidence scores of the other classifiers in the hierarchical classification structure by applying a confusion factor to the confidence score based on a correspondence between the confidence score and confidence scores of mutually exclusive classifiers in the hierarchical classification structure;

fifth instructions for generating a representation of the input media data based on the modified confidence score being higher than a predetermined threshold confidence score; and

sixth instruction for storing the generated representation of the input media data.

10. The computer program product of claim 9, wherein the fourth instructions for modifying the confidence score automatically for a classifier include instructions for boosting the confidence score using a boosting factor determined based on a correspondence between the confidence score and confidence scores of ancestor classifiers in the hierarchical classification structure.

11. The computer program product of claim 10, wherein the boosting factor is determined as a function of a difference between the confidence score of the classifier and confidence scores of each ancestor classifier in the hierarchical classification structure.

12. The computer program product of claim 9, wherein the fifth instructions for generating a representation of the input media data include instructions for generating an index data structure based on the modified confidence score.

13. The computer program product of claim 9, further comprising:

seventh instructions for using the stored representation of the input media data to access the input media data.

14. The computer program product of claim 13, wherein the seventh instructions for using the stored representation of the input media data to access the input media data includes:

instructions for receiving a search query;

instructions for determining if terms in the search query match terms in the stored representation of the input media data; and

instructions for accessing the input media data if the terms in the search query match terms in the stored representation of the input media data.

15. An apparatus for classifying media data, comprising:

means for classifying input media data for later access based on classification in response to receiving the input media data;

means for applying a plurality of classifiers in a hierarchical classification structure to the input media data;

means for generating confidence scores for the plurality of classifiers in the hierarchical classification structure;

means for modifying a confidence score automatically for a classifier in the hierarchical classification structure based on a relationship between the classifier and other classifiers in the hierarchical classification structure and the confidence scores of the other classifiers in the hierarchical classification structure by applying a confusion factor to the confidence score based on a correspondence between the confidence score and confidence scores of mutually exclusive classifiers in the hierarchical classification structure;

means for generating a representation of the input media data based on the modified confidence score being higher than a predetermined threshold confidence score; and

means for storing the generated representation of the input media data in a storage medium.

16. The apparatus of claim 15, wherein the means for modifying the confidence score automatically for a classifier includes means for boosting the confidence score using a boosting factor determined based on a correspondence between the confidence score and confidence scores of ancestor classifiers in the hierarchical classification structure.

17. The apparatus of claim 16, wherein the boosting factor is determined as a function of a difference between the confidence score of the classifier and confidence scores of each ancestor classifier in the hierarchical classification structure, and wherein the confusion factor is determined as a function of a difference between the confidence score of the classifier and a maximum confidence score from mutually exclusive classifiers in the hierarchical classification structure.