Discovering authoritative images of people entities

• US 10,509,963 B2
• Filed: 12/20/2012
• Issued: 12/17/2019
• Est. Priority Date: 12/20/2012
• Status: Active Grant

• Current Assignee: Microsoft Technology Licensing LLC (Microsoft Corporation)
• Original Assignee: Microsoft Technology Licensing LLC (Microsoft Corporation)
• Inventors: Kaheel, Ayman Malek Abdel Hamid, Gaggara, Padma Priya, Arul, Prakash Asirvatham, Hafeez, Mohammad Adil, Kulkarni, Dhananjay Dilip, Cao, Kancheng
• Primary Examiner(s): Ruiz, Angelica
• Application Number:
  US13/722,085
• Publication Number:
  US 20140177966A1
• Time in Patent Office: 2,553 Days
• Field of Search: 707600-831, 707899, 707999001-999206
• CPC Class Codes: G06F 16/951 Indexing; Web crawling techniques
• CPC Class Codes: G06F 16/9535 Search customisation based on user profiles and personalisation
• CPC Class Codes: G06V 20/30 in albums, collections or shared content, e.g. social network photos or video

Abstract

Systems, methods, and computer storage media for discovering authoritative images of people entities are provided. Selections of person entities are received. Authoritative URLs and authoritative images for the person entities are identified. Once the authoritative images are identified, features are extracted. Queries for the person entities are identified by mining search engine logs. The queries and features can be utilized to construct candidate queries to identify and retrieve candidate image URLs. Candidate features are extracted for each candidate image associated with the candidate image URLs. Training data may be utilized to train a classifier that can be run on each candidate image. Each candidate image can then be tagged with an entity ID tag. Images with the entity ID tag can be ranked higher in search engine results page than images without the entity ID tag.

Claims

1. A computer-implemented method of identifying authoritative images of people entities, the method comprising:

receiving a selection of a person entity;

identifying an authoritative image of the person entity;

extracting a first set of one or more features from the authoritative image of the person entity;

receiving, from search engine logs, queries used to search for the person entity;

constructing a candidate query based on the first set of one or more features from the authoritative image of the person entity and on the queries used to search for the person entity;

receiving a candidate image based on the candidate query being issued to a search engine image index;

extracting a second set of one or more features from the candidate image of the person entity;

determining a match based on comparing the first set of one or more features from the authoritative image of the person entity to the second set of one or more features from the candidate image of the person entity;

based on determining the match exists, tagging the candidate image with an entity ID tag, wherein the candidate image tagged with the entity ID tag causes a ranking of the candidate image tagged with the entity ID tag to be ranked higher in one or more search results of the person entity than of an image without the entity ID tag.

2. The method of claim 1, wherein the identifying authoritative image of the person entity includes identifying URLs from Wikipedia, IMDB page, and/or an academic reference.

3. The method of claim 2, wherein extracting a first set of one or more features from the authoritative image of the person entity includes the extracting of text features, wherein the text features include text associated with URLs associated with the authoritative image.

4. The method of claim 3, wherein extracting a first set of one or more features from the authoritative image of the person entity includes the extracting of click features.

5. The method of claim 4, wherein extracting a first set of one or more features from the authoritative image of the person entity includes the extracting of visual features.

6. The method of claim 5, wherein extracting a first set of one or more features from the authoritative image of the person entity includes the extracting of text similarity features.

7. The method of claim 5, wherein extracting a first set of one or more features from the authoritative image of the person entity includes the extracting of absolute rank scores, wherein the absolute rank scores indicate how many times an image is responsive to a query associated with a person entity.

8. The method of claim 7, wherein extracting a first set of one or more features from the candidate image of the person entity includes the extracting of text features, wherein the text features includes text associated with URLs associated with the candidate image.

9. The method of claim 5, wherein extracting a first set of one or more features from the candidate image of the person entity includes the extracting of absolute rank scores, wherein the absolute rank scores indicate how many times an image is responsive to a query associated with a person entity.

10. The method of claim 8, wherein the entity ID tag identifies the person entity associated with the image.

20. The method of claim 1, wherein the match includes varying degrees of the match.

11. A method of identifying and ranking one or more images of people entities, the method comprising:

identifying one or more images for a person entity from one or more authoritative URLs having a high confidence factor;

extracting a first set of one or more features from each of the one or more images, wherein the first set of one or more features from each of the one or more images includes absolute rank scores indicating a number of times each image is responsive to a query associated with the person entity;

mining search engine logs to identify queries for the person entity;

utilizing the queries and the first set of one or more features from each of the one or more images to construct candidate queries for the person entity;

issuing the candidate queries against an image index to retrieve candidate image URLs;

extracting a second set of one or more features from each candidate image of the person entity associated with the candidate image URLs;

determining a match based on comparing the first set of one or more features from each of the one or more images of the person entity to the second set of one or more features from each of the candidate image of the person entity;

based on determining the match exists, tagging the each candidate image with an entity ID tag; and

ranking images with the entity ID tag causing a ranking of the candidate image tagged with the entity ID tag to be ranked higher in one or more search engine results page than images without the entity ID tag.

12. The method of claim 11, wherein the first set of one or more features and/or the second set of one or more features further include text similarity features, click features, and visual features.

13. The method of claim 12, wherein visual features include face similarity features.

14. The method of claim 11, further comprising receiving a label associated with a set of training images.

15. The method of claim 14, wherein the label includes varying degrees of match or no match.

16. A computer system that facilitates discovering images of people entities, the computer system comprising a processor coupled to a computer storage medium, the computer storage medium having stored thereon a plurality of computer software components executable by the processor, the computer software components comprising:

a selection component that receives a selection of a person entity;

an entity extraction component that extracts similar entities to the person entity;

a uniform resource locator component that identifies authoritative uniform resource locators (URLs) for the person entity, wherein the authoritative URLs are determined to be trustworthy sources;

an image component that identifies images from the authoritative URLs for the person entity;

an authoritative feature extraction component that extracts a first set of one or more features from the images;

a mine component that mines search engine logs to identify queries for the person entity;

a construct component that utilizes the queries and the first set of one or more features to construct candidate queries for the person entity;

an issue component that issues the candidate queries against a search engine image index to retrieve candidate image URLs;

a candidate feature extraction component that extracts a second set of one or more candidate features from each candidate image from the candidate image URLs, wherein the candidate features include a number of times a candidate URL appears in different queries associated with the person entity;

classifier component determines a match based on comparing the first set of one or more features for the images from each of the one or more images of the person entity to the second set of one or more candidate features from each candidate image of the person entity;

based on determining the match exists, a tag component that tags each candidate image with an entity ID tag, and

a ranking component that ranks images with the entity ID tag causing a ranking of the candidate image tagged with the entity ID tag to be ranked higher in one or more search engine results page than images without the entity ID tag.

17. The system of claim 16, further comprising a training component that trains a classifier with training data associated with the first set of one or more features and/or the second set of one or more features.

18. The system of claim 17, wherein the classifier component runs the classifier on an image index.

19. The system of claim 18, wherein the entity ID tag identifies the person entity associated with the image.