Scalable metadata extraction for video search

• US 8,645,354 B2
• Filed: 06/23/2011
• Issued: 02/04/2014
• Est. Priority Date: 06/23/2011
• Status: Active Grant

• Current Assignee: Microsoft Technology Licensing LLC (Microsoft Corporation)
• Original Assignee: Microsoft Corporation
• Inventors: Kong, Xiao, Wang, Wei, Cai, Rui, Li, Haifeng, Sun, Yanfeng
• Primary Examiner(s): Fleurantin, Jean B
• Assistant Examiner(s): Ly, Anh
• Application Number:
  US13/166,813
• Publication Number:
  US 20120330952A1
• Time in Patent Office: 957 Days
• Field of Search: 707/709, 707/769, 707/748, 707/E17.014, 707/E17.108, 709/217, 715/760, 715/771, 715/786, 715/234, 725/115, 725/135, 725/143, 725/145, 725/41, 725/43, 725/44
• US Class Current: 707/709
• CPC Class Codes: G06F 16/7867 using information manually generated, e.g. tags, keywords, comments, title and artist information, manually generated time, location and usage information, user ratings
• CPC Class Codes: G06F 16/951 Indexing; Web crawling techniques
• CPC Class Codes: G06F 16/9538 Presentation of query results

Abstract

Video entity templates defining common features that relate to various metadata types shared among a group of video Web pages are generated for target Web sites. Metadata associated with videos contained within Web pages belonging to a particular target Web site can then be automatically and accurately extracted using a video entity template generated for the particular target Web site. This metadata can then be indexed for use by video search applications in providing video search results.

Claims

1. A computer-implemented method for extracting metadata, the computer-implemented method comprising performing computer-implemented operations for:

grouping, by a computer comprising one or more processors, a plurality of Web pages into a group based upon a common visual layout shared among the plurality of Web pages, wherein the group is one of a plurality of groups formed according to common visual layout shared among additional Web pages considered for grouping;

removing, by the computer, one or more regions shared among the plurality of Web pages from each of the plurality of Web pages in the group to define a candidate region, and removing one or more regions each containing common elements shared among the plurality of Web pages in the group;

extracting, by the computer, one or more candidate features from the candidate region; and

selecting, by the computer, one of the one or more candidate features for use in a video entity template.

2. The computer-implemented method of claim 1, wherein each of the plurality of Web pages comprises a video.

3. The computer-implemented method of claim 2, wherein the group is selected as a target group.

4. The computer-implemented method of claim 2, wherein removing the one or more regions shared among the plurality of Web pages from each of the plurality of Web pages in the group to define the candidate region comprises removing one or more regions each comprising a repeat region shared among the plurality of Web pages in the group.

5. The computer-implemented method of claim 2, wherein extracting the one or more candidate features from the candidate region comprises extracting the one or more candidate features from the candidate region based upon one or more attributes specific to a particular entity.

6. The computer-implemented method of claim 5, wherein the particular entity is a title entity, and the one or more candidate features share the one or more attributes specific to the title entity.

7. The computer-implemented method of claim 5, wherein the particular entity is a publication date entity, and the one or more candidate features share the one or more attributes specific to the publication date entity.

8. The computer-implemented method of claim 5, wherein the particular entity is a view count entity, and the one or more candidate features share the one or more attributes specific to the view count entity.

9. The computer-implemented method of claim 2, further comprising cross-validating the one of the one or more candidate features that was selected with other candidates from other pages within the group.

10. A computer-implemented method for extracting metadata, the computer-implemented method comprising performing computer-implemented operations for:

generating, by a computer comprising one or more processors, a page group comprising a plurality of Web pages, each of which contains a video, wherein the page group is one of a plurality of page groups formed according to common visual layout shared among additional web pages considered for grouping;

removing one or more regions shared among the plurality of Web pages from each of the plurality of Web pages in the page group to define a candidate region, and removing one or more regions each containing common elements shared among the plurality of Web pages in the page group;

generating, by the computer, a video entity template from the page group, the video entity template comprising identification of an entity;

matching, by the computer, a target page to the video entity template;

extracting, by the computer, metadata associated with the entity from the target page utilizing the video entity template; and

indexing, by the computer, the metadata extracted from the target page in a video search index.

11. The computer-implemented method of claim 10, wherein the plurality of Web pages are associated with a single Web site.

12. The computer-implemented method of claim 10, wherein generating the video entity template from the page group comprises:

removing one or more duplicate elements shared among the plurality of Web pages;

removing one or more repeat regions shared among the plurality of Web pages;

extracting one or more candidate features from a resultant region of each of the plurality of Web pages based upon one or more attributes of the entity;

selecting one of the one or more candidate features from the resultant region to be included in the video entity template; and

including the one candidate features selected from the one or more candidate features in the video entity template.

13. The computer-implemented method of claim 12, wherein extracting the one or more candidate features based upon the one or more attributes of the entity comprises extracting one or more HTML features associated with a title of each of the plurality of Web pages.

14. The computer-implemented method of claim 12, wherein extracting the one or more candidate features based upon the one or more attributes of the entity comprises extracting one or more dates displayed in a particular date format.

15. The computer-implemented method of claim 12, wherein extracting the one or more candidate features based upon the one or more attributes of the entity comprises extracting one or more numbers displayed proximate to a white list term.

16. A computer storage medium that does not include signals having computer readable instructions stored thereupon that, when executed by a computer, cause the computer to:

group Web pages of a Web site by visual layout into a plurality of groups, wherein each group of the plurality of groups is one of a plurality of groups formed according to common visual layout shared among additional Web pages considered for grouping;

remove one or more regions shared among the Web pages from each of the Web pages in the plurality of groups to define a candidate region for each group of the plurality of groups, and removing one or more regions each containing common elements shared among the Web pages in each group;

select a target group from the plurality of groups from which to generate a video entity template;

remove common elements of the Web pages in the target group;

remove repeat regions of the Web pages in the target group;

extract one or more candidate features from a remaining candidate region for the Web pages in the target group, the one or more candidate features being candidate features for a particular target entity;

select a particular candidate feature of the one or more candidate features for the particular target entity;

cross-validate the particular candidate feature to previously selected candidate features from one or more other groups;

if cross-validation fails, return to candidate feature extraction;

if cross-validation is successful, generate the video entity template; and

output the video entity template.

17. The computer storage medium of claim 16, further comprising instructions that, when executed by the computer, cause the computer to incorporate the video entity template into a video search index.

18. The computer storage medium of claim 16, further comprising instructions that, when executed by the computer, cause the computer to receive a target Web page and extract metadata associated with the particular target entity utilizing the video entity template.