Method and system for tagging objects comprising tag recommendation based on query-based ranking and annotation relationships between objects and tags

• US 9,116,894 B2
• Filed: 03/14/2013
• Issued: 08/25/2015
• Est. Priority Date: 03/14/2013
• Status: Active Grant

• Current Assignee: Xerox Corporation (Xerox Holdings Corp.)
• Original Assignee: Xerox Corporation (Xerox Holdings Corp.)
• Inventors: Chidlovskii, Boris
• Primary Examiner(s): Truong, Dennis
• Application Number:
  US13/828,048
• Publication Number:
  US 20140280232A1
• Time in Patent Office: 894 Days
• Field of Search: 707/737, 707/769, 707/802, 707/749, 382/159, 382/190, 382/224, 715/230
• US Class Current: 1/1
• CPC Class Codes: G06F 16/48 Retrieval characterised by using metadata, e.g. metadata not derived from the content or metadata generated manually
• CPC Class Codes: G06F 16/58 Retrieval characterised by using metadata, e.g. metadata not derived from the content or metadata generated manually

Abstract

A method and system is disclosed for tagging a latent object with selected tag recommendations, including a set of content objects wherein each object is characterized by an associated set of content features. An annotation relationship is determined between the features and a pre-determined tag for the each object, the relationship being defined by a graph construction representative of an affinity relationship between each pre-selected tag and content object to a selected query. A plurality of the annotation relationships are ranked based upon a relevance of the preselected tags to the content features in response to a new query for assigning a new tag to the each object, so that a suggested tag is made from the ranking whereby the suggested tag is determined as a most likely tag for annotating the content object.

Claims

1. A method for tagging a latent object with selected tag recommendations, including:

receiving an input from a user for tagging the latent object wherein the input includes a user tag preference and wherein the latent object is characterized by a set of predetermined tags representative of an associated set of content features;

generating a query using the tag preference for comparing the tag preference to the set of predetermined tags;

determining a first annotation relationship between the features and the set of predetermined tags for the object, the relationship being defined by a graph construction representative of an affinity relationship between each predetermined tag and the object content features;

determining a second annotation relationship representative of frequency of tagging usage of each of the set of predetermined tags;

ranking the first and second annotation relationships based upon a weighted relevance of the predetermined tags and the user tag preference to the object content features using a neighborhood linearization technique to infer edge weights; and

suggesting a plurality of suggested tags from the ranking whereby the suggested tags are determined as most likely for annotating the content object.

2. The method of claim 1 wherein the determining the first annotation relationship comprises constructing a sparse affinity graph representative of the affinity relationship.

3. The method of claim 1 wherein the ranking comprises computing ranking vectors for the content objects and predetermined tags, respectively, in accordance with an iterative evaluation for identifying a converging correlation between the content objects and the predetermined tags in response to the query.

4. The method of claim 1 wherein the determining the first annotation relationship comprises an integrating of the annotation relationship between the latent object, the predetermined tags, and the user tag preference.

5. The method of claim 1 wherein the defining of the graph construction comprises a use of object features and an l₁-norm minimization.

6. The method of claim 5 further including determining an affinity relationship between the predetermined tags using a sparse graph construction based on l₁-norm minimization for building a tag affinity graph.

7. The method of claim 1 wherein the ranking includes deploying a Laplacian regularization framework on the graph construction for assessing a label propagation framework.

8. The method of claim 7 wherein the graph construction comprises object and tag affinity graphs.

9. The method of claim 1 wherein the determining and ranking comprise latent objects including images or documents.

19. The method of claim 1 further including comparing the ranking of the first annotation relationship and the ranking of the second annotation relationship to the user tag preference.

10. A tag recommendation system for annotating content objects including:

an annotation module comprising an annotation relationship detector and ranking processor wherein each object is characterized by an associated set of content features and the processor determines a first annotation relationship between the features and predetermined tags for the object, the relationship being defined by a graph construction representative of an affinity relationship between the predetermined tags and content objects to a selected query; and

,

determining a second annotation relationship representative of frequency of tagging usaqe of each of the set of predetermined tags, wherein the processor ranks a plurality of the annotation relationships based upon a relevance of the predetermined tags to the content features in response to a user input of a user tag preference for assigning a new tag to the object, and suggesting suggested tags from the ranking as most likely for annotating the content object.

11. The system of claim 10 wherein the first annotation relationship comprises a construction of a sparse affinity graph representative of the affinity relationship.

12. The system of claim 11 wherein the graph construction comprises a use of object features and an l₁-norm minimization.

13. The system of claim 10 wherein the first annotation relationship comprises computing ranking vectors for the content objects and predetermined tags, respectively, in accordance with an iterative evaluation for identifying a converging correlation between the content objects and the predetermined tags in response to the selected query.

14. The system of claim 10 wherein the annotation relationship comprises an integration of the annotation relationship between the latent object and the predetermined tags.

15. The system of claim 10 wherein the first annotation relationship comprises a Laplacian regularization framework on the graph construction for assessing a label propagation framework.

16. The system of claim 15 wherein the graph construction comprises object and tag affinity graphs.

17. The system of claim 10 wherein the content objects are internet accessible.

18. The system of claim 10 wherein the content objects comprise a computer memory stored collection of content objects.

20. The method of claim 10 further including comparing the ranking of the first annotation relationship and the ranking of the second annotation relationship to the user tag preference.