Personalized tag ranking

• US 9,087,111 B2
• Filed: 09/12/2012
• Issued: 07/21/2015
• Est. Priority Date: 01/18/2010
• Status: Expired due to Fees

• Current Assignee: International Business Machines Corporation
• Original Assignee: International Business Machines Corporation
• Inventors: Mass, Yosi, Shmueli-Scheuer, Michal
• Primary Examiner(s): Drennan, Barry
• Assistant Examiner(s): Woldemariam, Aklilu
• Application Number:
  US13/612,698
• Publication Number:
  US 20130016916A1
• Time in Patent Office: 1,042 Days
• Field of Search: 382/305, 382/220, 382/118, 382/115, 382/117, 382/224, 707/3, 707/5, 725/28, 725/34, 725/35
• US Class Current: 1/1
• CPC Class Codes: G06F 11/1471 involving logging of persistent data for recovery
• CPC Class Codes: G06F 16/254 Extract, transform and load [ETL] procedures, e.g. ETL data flows in data warehouses
• CPC Class Codes: G06F 16/258 Data format conversion from or to a database
• CPC Class Codes: G06F 16/58 Retrieval characterised by using metadata, e.g. metadata not derived from the content or metadata generated manually
• CPC Class Codes: G06F 2201/84 Using snapshots, i.e. a logical point-in-time copy of the data

Abstract

Personalized tag ranking of images, including identifying within a reference image collection any images that are similar to an input image, identifying within a source image collection any images that have associated tags that are similar to a set of input tags associated with the input image, identifying among the images identified in the reference image collection any images that are similar to the images identified in the source image collection, and calculating a weight for each of a plurality of tag pairs, where each of the tags in each of the tag pairs is associated with a different subset of the images in the reference image collection identified as being similar to the images identified in the source image collection, and ranking the input tags of the input image in accordance with a predefined ranking function as applied to the tag pair weights.

Claims

1. A method for personalized tag ranking of images, the method comprising:

identifying within a reference image collection any images that are similar to an input image in accordance with a first predefined image similarity measurement;

identifying within a source image collection any images that have associated tags that are similar to a set of input tags in accordance with a predefined tag similarity measurement, wherein said set of input tags is associated with said input image;

identifying among said images identified in said reference image collection any images that are similar to said images identified in said source image collection in accordance with a second predefined image similarity measurement;

calculating a weight for each of a plurality of tag pairs, wherein each of said tags in each of said tag pairs is associated with a different subset of said images in said reference image collection identified as being similar to said images identified in said source image collection, and wherein said weight is calculated in accordance with a predefined weight function; and

ranking said input tags of said input image in accordance with a predefined ranking function as applied to said tag pair weights.

2. The method according to claim 1 and further comprising representing said images identified in said source image collection in a plurality of groups, wherein each of said groups is associated with a different tag in said set of input tags.

3. The method according to claim 2 and further comprising representing in a plurality of subsets of said groups said images in said reference image collection identified as being similar to said images identified in said source image collection, wherein each of said subsets is associated with said tag of said group of which said subset is a subset.

4. The method according to claim 3 and further comprising calculating a similarity value between said images represented by a pair of said subsets in accordance with a third predefined image similarity measurement.

5. The method according to claim 4 and further comprising calculating said weight wherein said predefined weight function is applied to said similarity value and a tag-based similarity value calculated between said tags associated with said pair of subsets.

6. The method according to claim 1 and further comprising:

building a tag graph of a plurality of nodes, wherein each node in said tag graph corresponds to a different tag in said set of input tags; and

assigning each of said weights to each of a plurality of edges of said tag graph corresponding to one of said tag pairs.

11. The method according to claim 1 and further comprising performing said identifying, comparing, and ranking wherein each of said input tags includes at least one word describing a characteristic of said input image.

7. A method for personalized tag ranking of images, the method comprising:

identifying within a reference image collection any images that are similar to an input image in accordance with a first predefined image similarity measurement;

identifying within a source image collection any images that have associated tags that are similar to a set of input tags in accordance with a predefined tag similarity measurement, wherein said set of input tags is associated with said input image;

representing in a plurality of groups said images identified within said source image collection, wherein each of said groups is associated with a different tag in said set of input tags;

identifying within said groups of images identified within said source image collection any images that are similar to said images identified within said reference image collection in accordance with a second predefined image similarity measurement;

representing in a plurality of subsets of said groups said images within said source image collection identified as being similar to said images identified within said reference image collection, wherein each of said subsets is associated with said tag of said group of which said subset is a subset;

calculating for any pair of said subsets

a similarity value between said images represented by said pair of subsets in accordance with a third predefined similarity measurement, and

a weight for each pair of said tags associated with said pair of subsets as a predefined function of said similarity value and a predefined function of a tag-based similarity value calculated between said pair of tags; and

ranking said input tags of said input image in accordance with a predefined ranking function as applied to said tag pair weights.

8. The method according to claim 7 and further comprising:

building a tag graph of a plurality of nodes, wherein each node in said tag graph preferably corresponds to a different tag in said set of input tags; and

assigning each of said weights to a different edge of said tag graph corresponding to a different one of said pair of said tags.

9. The method according to claim 7 wherein said step of identifying within said reference image collection comprises performing a content-based search using said input image as a query image

10. The method according to claim 7 wherein said step of identifying within said reference image collection comprises identifying where said reference image collection includes a plurality of images that relate to the individual interests of a user.

12. The method according to claim 7 and further comprising performing said identifying, representing, calculating, and ranking wherein each of said input tags includes at least one word describing a characteristic of said input image.