Automated propagation of document metadata

• US 7,107,520 B2
• Filed: 11/18/2002
• Issued: 09/12/2006
• Est. Priority Date: 11/18/2002
• Status: Active Grant

• Current Assignee: Hewlett-Packard Development Company, L.P. (HP Inc.)
• Original Assignee: Hewlett-Packard Development Company, L.P. (HP Inc.)
• Inventors: Gargi, Ullas
• Primary Examiner(s): Bashore, William
• Assistant Examiner(s): NGUYEN, CHAU T
• Application Number:
  US10/298,919
• Publication Number:
  US 20040098362A1
• Time in Patent Office: 1,394 Days
• Field of Search: 715/512, 715/514, 715529-530, 715/538, 707/102, 707/104.1
• US Class Current: 715/210
• CPC Class Codes: G06F 16/51 Indexing; Data structures therefor; Storage structures

Abstract

An automated method for conditionally propagating metadata instances among documents includes defining groups of metadata instances and includes assigning different propagation coefficients to each group. Each propagation coefficient assigned to a particular group is based upon correlating the associated document attribute or attributes with reliable propagations of metadata instances within the group. After the groups of metadata instances have been defined and the propagation coefficients have been assigned, propagation of a metadata instance from a second document to a first document is based upon a combination of (1) determining the particular group in which the metadata instance resides, (2) identifying the propagation coefficients that are assigned to that group for the associated document attributes, and (3) comparing the first and second documents with respect to the associated document attributes.

Claims

1. A computer-implemented method of conditionally propagating metadata among documents comprising:

defining a plurality of groups of metadata instances;

assigning a plurality of propagation coefficients to each said group, where each said propagation coefficient assigned to each said group is associated with at least one document attribute and is based upon correlating said associated document attribute with determinations of document-to-document propagations of said metadata instances of said group;

selecting first and second documents between which a specific metadata instance may be propagated, said first and second documents being stored in memory;

basing metadata propagation from said second document to said first document upon a plurality of said document attributes, including

(a) determining a particular said group in which said specific metadata instance resides;

(b) for each of said document attributes upon which said metadata propagation is based, identifying said associated propagation coefficient that was assigned to said particular group; and

(c) for each of said document attributes upon which said metadata propagation is based;

comparing said first and second documents with respect to said document attributes; and

on a basis of said document attributes, selectively propagating said specific metadata instance to said first document for storage in said memory.

2. The computer-implemented method of claim 1 wherein basing said metadata propagation to said first document upon said plurality of document attributes includes calculating a metadata weighting that is determined collectively by said associated propagation coefficients that are assigned to said particular group and by degrees of similarity in comparing said first and second documents with respect to said document attributes.

3. The computer-implemented method of claim 2 wherein calculating said metadata weighting is further determined by a source weighting assigned to said specific metadata instance as applied to said second document, said source weighting being representative of a confidence that said specific metadata instance is accurately applied to said second document.

4. The computer-implemented method of claim 3 further comprising associating said metadata weighting with said first document when said specific metadata instance is propagated to said first document, said metadata weighting being representative of a confidence that said specific metadata instance is accurately applied to said first document.

5. The computer-implemented method of claim 4 wherein calculating said metadata weighting includes limiting said calculation such that said metadata weighting does not exceed said source weighting.

6. The computer-implemented method of claim 3 further comprising calculating said source weighting to be responsive to a number of propagation occurrences that said specific metadata instance experienced in being applied to said second document, said source weighting being reduced at each said propagation occurrence so as to be representative of a lowered confidence.

7. The computer-implemented method of claim 2 wherein calculating said metadata weighting is based on steps (a), (b), and (c) and includes a summing of the various produce of said propagation coefficients times the similarity levels that said first and second documents have with respect to said document attributes with which said propagation coefficients are associated.

8. The computer-implemented method of claim 1 wherein said selecting said first and second documents is from a common storage of documents and wherein said metadata propagation relates to a human-generated metadata instance attached to said second document.

9. The computer-implemented method of claim 8 wherein said step of defining said groups includes identifying groups of related annotations attached to said documents, said groups including a first group that relates to personal names and a second group that relates to geographical locations.

10. The computer-implemented method of claim 1 wherein basing said metadata propagation upon said plurality of document attributes includes referencing:

(a) document access and usage attributes that identify access patterns of said first and second documents;

(b) intrinsic non-content attributes that identify aspects of said first and second documents without directly determining content of said first and second documents; and

(c) intrinsic content-specific attributes that relate directly to said content of said first and second documents.

11. The computer-implemented method of claim 10 further comprising enabling a user to customize the relative weights assigned to different said attributes with regard to executing said metadata propagation.

12. A computer-implemented method for propagating metadata instances of a particular metadata type comprising:

grouping said metadata instances in accordance with similarities among corresponding document attributes of training documents with which said metadata instances are associated;

for each grouping of metadata instances, assigning correlation coefficients to relevant said document attributes, where said correlation coefficients are computed based on correlations between said grouping of metadata instances and individual said relevant document attributes; and

propagating specific metadata instances to various second documents based on said groupings and on said correlation coefficients, said propagating being further based upon;

(a) source weightings of said specific metadata instances as applied to source documents from which said specific metadata instances are to be propagated, said source weightings being representative of a perceived confidence in attachments of said specific metadata instances to said source documents; and

(b) calculating a metadata weighting when a particular metadata instance is propagated from a source document to a particular said second document, said metadata weighting representing factors that include similarities between said source document and said particular second document, wherein calculating said metadata weighting (m_y^g) when said particular metadata instance is propagated from said source document (x) to said particular said second document (y) includes factoring;

(1) said source weighting (m_x^g) of the source document from which said particular metadata instance is propagated;

(2) said correlation coefficients for said grouping (g) in which said particular metadata instance resides, said correlation coefficients being propagation coefficients (α

(c, g)) for said relevant document attributes (c) selected for said grouping; and

(3) similarities (S_c(x, y)) between said source and said particular said second document with respect to said relevant document attributes selected for said grouping.

13. The computer-implemented method of claim 12 wherein said calculating is implemented using the propagation equation:

14. The computer-implemented method of claim 12 wherein said calculating further includes providing a reduction on a basis of a number of times said particular metadata instance has been propagated from an original source document.

15. The computer-implemented method of claim 12 wherein said grouping includes using said training documents as a set of documents for defining metadata groups and determining said propagation coefficients.

16. The computer-implemented method of claim 15 wherein using said training documents includes forming a propagation database that identifies (1) said metadata groups, (2) said metadata instances in each said metadata group, and (3) each said propagation coefficient assigned to said metadata instances.