Generating a taxonomy for documents from tag data

1. A method, comprising:

clustering tags on documents using tag weightings of the tags on the documents, where each cluster comprises an identified subject;

comparing the identified subjects to identify relationships between the identified subjects; and

building a taxonomy of subjects using the identified relationships between the identified subjects programmatically without user intervention.

2. The method of claim 1 , where clustering the tags on the documents using the tag weightings of the tags on the documents, where each cluster comprises the identified subject, comprises:

processing the tags on the documents using a non-overlapping clustering technique; and

using tag similarity and co-occurrence of the tags to cluster the tags into the identified subjects.

3. The method of claim 1 , further comprising:

determining an association of each tag with the documents;

representing each association of each tag with the documents as a vector of a length representative of a number of the documents with which the tag is associated, with each element of the vector representing the respective tag weighting of the tag on the documents;

correlating pairs of vectors; and

determining a similarity between the tags based upon the correlation of each pair of vectors.

4. The method of claim 3 , where clustering the tags on the documents using the tag weightings of the tags on the documents, where each cluster comprises the identified subject, comprises clustering the tags on the documents based upon the determined similarity between the tags based upon the correlation of each pair of vectors.

5. The method of claim 1 , where building the taxonomy of subjects using the identified relationships between the identified subjects programmatically without user intervention comprises building a hierarchical taxonomy of subjects, where higher-level subject terms within the hierarchical taxonomy of subjects represent more-generic subjects and classify a wider-range of documents relative to lower-level subject terms within the hierarchical taxonomy of subjects.

6. The method of claim 1 , further comprising creating a graph with the documents as nodes and edges between the nodes representing a determined similarity between the documents.

7. The method of claim 1 , where each of the tag weightings of the tags on the documents comprise a combination of a number of instances of the tag being applied to at least one document with at least one of a positive rating and a negative rating of the tag in the at least one document.

8. A system, comprising:

a memory; and

a processor programmed to:

cluster tags on documents using tag weightings of the tags on the documents, where each cluster comprises an identified subject;

compare the identified subjects to identify relationships between the identified subjects; and

build a taxonomy of subjects within the memory using the identified relationships between the identified subjects programmatically without user intervention.

9. The system of claim 8 , where, in being programmed to cluster the tags on the documents using the tag weightings of the tags on the documents, where each cluster comprises the identified subject, the processor is programmed to:

process the tags on the documents using a non-overlapping clustering technique; and

use tag similarity and co-occurrence of the tags to cluster the tags into the identified subjects.

10. The system of claim 8 , where the processor is further programmed to:

determine an association of each tag with the documents;

represent each association of each tag with the documents as a vector of a length representative of a number of the documents with which the tag is associated, with each element of the vector representing the respective tag weighting of the tag on the documents;

correlate pairs of vectors; and

determine a similarity between the tags based upon the correlation of each pair of vectors.

11. The system of claim 10 , where, in being programmed to cluster the tags on the documents using the tag weightings of the tags on the documents, where each cluster comprises the identified subject, the processor is programmed to cluster the tags on the documents based upon the determined similarity between the tags based upon the correlation of each pair of vectors.

12. The system of claim 8 , where, in being programmed to build the taxonomy of subjects within the memory using the identified relationships between the identified subjects programmatically without user intervention, the processor is programmed to build a hierarchical taxonomy of subjects, where higher-level subject terms within the hierarchical taxonomy of subjects represent more-generic subjects and classify a wider-range of documents relative to lower-level subject terms within the hierarchical taxonomy of subjects.

13. The system of claim 8 , where the processor is further programmed to create a graph with the documents as nodes and edges between the nodes representing a determined similarity between the documents.

14. The system of claim 8 , where each of the tag weightings of the tags on the documents comprise a combination of a number of instances of the tag being applied to at least one document with at least one of a positive rating and a negative rating of the tag in the at least one document.

15. A computer program product, comprising:

a computer-readable storage medium having computer-readable program code embodied therewith comprising computer-executable instructions where the computer-executable instructions when executed by a computer cause the computer to:

cluster tags on documents using tag weightings of the tags on the documents, where each cluster comprises an identified subject;

compare the identified subjects to identify relationships between the identified subjects; and

build a taxonomy of subjects using the identified relationships between the identified subjects programmatically without user intervention.

16. The computer program product of claim 15 , where, in causing the computer to cluster the tags on the documents using the tag weightings of the tags on the documents, where each cluster comprises the identified subject, the computer-executable instructions when executed by the computer cause the computer to:

process the tags on the documents using a non-overlapping clustering technique; and

use tag similarity and co-occurrence of the tags to cluster the tags into the identified subjects.

17. The computer program product of claim 15 , where the computer-executable instructions when executed by the computer further cause the computer to:

determine an association of each tag with the documents;

represent each association of each tag with the documents as a vector of a length representative of a number of the documents with which the tag is associated, with each element of the vector representing the respective tag weighting of the tag on the documents;

correlate pairs of vectors; and

determine a similarity between the tags based upon the correlation of each pair of vectors.

18. The computer program product of claim 17 , where, in causing the computer to cluster the tags on the documents using the tag weightings of the tags on the documents, where each cluster comprises the identified subject, the computer-executable instructions when executed by the computer cause the computer to cluster the tags on the documents based upon the determined similarity between the tags based upon the correlation of each pair of vectors.

19. The computer program product of claim 15 , where, in causing the computer to build the taxonomy of subjects using the identified relationships between the identified subjects programmatically without user intervention, the computer-executable instructions when executed by the computer cause the computer to build a hierarchical taxonomy of subjects, where higher-level subject terms within the hierarchical taxonomy of subjects represent more-generic subjects and classify a wider-range of documents relative to lower-level subject terms within the hierarchical taxonomy of subjects.

20. The computer program product of claim 15 , where the computer-executable instructions when executed by the computer further cause the computer to create a graph with the documents as nodes and edges between the nodes representing a determined similarity between the documents.

21. The computer program product of claim 15 , where each of the tag weightings of the tags on the documents comprise a combination of a number of instances of the tag being applied to at least one document with at least one of a positive rating and a negative rating of the tag in the at least one document.