UTILIZATION OF FEATURES EXTRACTED FROM STRUCTURED DOCUMENTS TO IMPROVE SEARCH RELEVANCE

• US 20130031032A1
• Filed: 07/27/2011
• Published: 01/31/2013
• Est. Priority Date: 07/27/2011
• Status: Active Grant

• Current Assignee: Microsoft Technology Licensing LLC (Microsoft Corporation)
• Original Assignee: Microsoft Corporation
• Inventors: Mehta, Rupesh Rasiklal, Nagaralu, Sree Hari, Das, Anjana, Mitra, Bhaskar
• Granted Patent: US 8,788,436 B2
• Time in Patent Office: Days
• US Class Current: 706/12
• CPC Class Codes: G06F 16/313 Selection or weighting of terms for indexing
• CPC Class Codes: G06F 16/80 of semi-structured data, e.g. markup language structured data such as SGML, XML or HTML content-based retrieval of web data G06F16/95
• CPC Class Codes: G06F 16/81 Indexing, e.g. XML tags; Data structures therefor; Storage structures
• CPC Class Codes: G06F 16/951 Indexing; Web crawling techniques
• CPC Class Codes: G06V 20/62 Text, e.g. of license plates, overlay texts or captions on TV images

Abstract

Features automatically extracted from semi-structured web pages are utilized by a search engine to rank documents that include semi-structured web pages. These features include, but are not limited to, a number of reviews, a number of positive reviews, and/or a number of negative reviews from a web page that includes user reviews. These features also include a number of views of a video that is viewable by way of a semi-structured web page. The features also include a number of subscribers to broadcasts of an individual from a social networking web page and a number of contacts of an individual listed on a social networking web page.

Claims

1. A method that facilitates ranking documents with respect to a received query, the method comprising:

at a general purpose search engine, receiving the query from a user that is configured to retrieve at least one document that is indexed by the general purpose search engine; and

causing a processor to output a ranked list of documents to the user based at least in part upon the query, the ranked list of documents comprising a semi-structured web page, a position of the semi-structured web page in the ranked list of documents based at least in part upon a value of a feature that is extracted from the semi-structured document at a learned location in the semi-structured document that is known to include the feature, wherein the position of the semi-structured document in the ranked list of documents is independent of any correlation between text of the query and the value of the feature.

2. The method of claim 1, wherein the semi-structured web page is a retailer page that describes a product, and wherein the feature is at least one of total number of reviews posted on the semi-structured web page by purchasers of the product, total number of positive reviews posted on the semi-structured web page by purchasers of the product, or total number of negative reviews posted on the semi-structured web page by purchasers of the product.

3. The method of claim 1, wherein the semi-structured web page is a web page that includes a video, and wherein the feature is a number of views of the video by viewers of the video on the web page.

4. The method of claim 1, wherein the semi-structured web page is a profile page of an entity on social networking site, and wherein the feature is a number of contacts of the entity.

5. The method of claim 1, wherein the semi-structured web page is a profile page of an entity on a social network site, and wherein the feature is a number of individuals that subscribe to broadcasts generated by the entity.

6. The method of claim 1, wherein content of the semi-structured web page fails to conform to structure of a relational database but comprises a pattern that is consistent across a plurality of other semi-structured web pages.

7. The method of claim 1, further comprising:

automatically extracting features from a plurality of semi-structured web pages, wherein the feature is automatically extracted from the semi-structured web page at the learned location.

8. The method of claim 7, wherein wrapper induction is employed to learn a wrapper that automatically or semi-automatically extracts the features from the plurality of semi-structured web pages.

9. The method of claim 8, further comprising:

learning a function that assigns scores to the feature, wherein a ranker of the general purpose search engine is trained based at least in part upon scores output by the function.

10. The method of claim 9, wherein the ranker comprises at least one decision tree.

11. A system that facilitates outputting a ranked list of search results responsive to receipt of a query, the system comprising:

a receiver component that receives the query from a user; and

a ranker component that outputs a ranked list of documents responsive to receipt of the query, the ranked list of documents comprising a semi-structured web page at a position amongst the ranked list of documents, the position amongst the ranked list of documents based at least in part upon a value of a feature that is at a learned location in the semi-structured web page, the position of the semi-structured web page amongst the ranked list of documents being independent of any correlation between the query and the value of the feature.

12. The system of claim 11 comprised by a general purpose search engine.

13. The system of claim 11, the ranker component comprising a computer-executable decision tree.

14. The system of claim 11, wherein the semi-structured web page is a web page that is configured for sale of a product, and wherein the feature is one of a number of reviews posted on the web page by purchasers of the product, a number of positive reviews posted on the web page by purchasers of the product, or a number of negative reviews posted on the web page by purchasers of the product.

15. The system of claim 11, wherein the semi-structured web page is a web page that is configured for sale of a product, and wherein the feature is one of a number of terms labeled as positive in reviews of the product posted on the web page by purchasers of the product or a number of terms labeled as negative in reviews of the product posted on the web page by purchasers of the product.

16. The system of claim 11, wherein the semi-structured web page is a web page that is configured to present a video to a user, and wherein the feature is a number of views of the video by way of the web page.

17. The system of claim 11, wherein the semi-structured web page is a web page that includes a profile of a particular individual, and wherein the feature is a number of contacts of the individual in a social networking web site.

18. The system of claim 11, wherein the semi-structured web page is a web page that includes a profile of a particular entity, and wherein the feature is a number of other individuals that have subscribed to receive broadcasts generated by the particular entity.

19. The system of claim 11, further comprising a learner component that learns a wrapper by way of wrapper induction, wherein the wrapper automatically extracts features from semi-structured web pages.

20. A computer-readable medium comprising instructions that, when executed by a processor, causes the processor to perform acts comprising:

receiving a query from a user;

extracting a value of a feature from a semi-structured web page independent of content of the query, the feature being one of;

a number of reviews posted on the semi-structured web page by purchasers of a product that is displayed on the web page;

a number of positive reviews posted on the semi-structured web page by purchasers of the product that is displayed on the web page;

a number of negative reviews posted on the semi-structured web page by purchasers of the product that is displayed on the web page;

a number of views of a video that is embedded on the web page;

a number of contacts of an entity whose profile is included on the web page;

a number of subscribers of an entity that broadcasts messages that are displayed on the web page;

responsive to receiving the query, providing to the user a ranked list of search results, the ranked list or search results comprising a plurality of documents displayed in a particular order, the plurality of documents comprising the semi-structured web page that is at a certain position in the particular order, wherein the certain position in the particular order is based at least in part upon the value of the feature extracted from the semi-structured web page, and wherein the certain position in the particular order is independent of any correlation between the value of the feature and the query.