Multi-domain recommendations

• US 8,504,583 B1
• Filed: 02/14/2012
• Issued: 08/06/2013
• Est. Priority Date: 02/14/2012
• Status: Active Grant

• Current Assignee: Microsoft Technology Licensing LLC (Microsoft Corporation)
• Original Assignee: Microsoft Corporation
• Inventors: Ke, Yan, Fan, Xiaodong, Li, Yang
• Primary Examiner(s): LE, HUNG D
• Application Number:
  US13/396,056
• Publication Number:
  US 20130212120A1
• Time in Patent Office: 539 Days
• Field of Search: None
• US Class Current: 707/767
• CPC Class Codes: G06F 16/9535 Search customisation based on user profiles and personalisation

Abstract

Methods, computer systems, and computer-readable storage media for generating and presenting multi-domain recommendations to a user viewing a Web page are provided. A topic-of-interest is identified on the Web page being viewed by the user. The topic-of-interest is mapped to an entity ID in a multi-domain recommendation data structure. The multi-domain recommendation data structure is used to identify entities, uniform resource locators, and search queries mapped to the entity ID. Multi-domain recommendations are generated using the entities, uniform resource locators, and search queries; the multi-domain recommendations are presented on the Web page being viewed by the user.

Claims

1. One or more computer-readable storage media having embodied thereon computer-executable instructions that, when executed, perform a method of generating and presenting multi-domain recommendations to a user viewing a topic-of-interest on a Web page, the method comprising:

identifying the topic-of-interest on the Web page being viewed by the user;

mapping the topic-of-interest to an entity ID in a multi-domain recommendation data structure stored in association with a data store;

using the multi-domain recommendation data structure;

(1) determining one or more entities mapped to the entity ID, wherein an entity comprises at least one of person, object, location, or topic;

(2) determining one or more uniform resource locators (URLs) mapped to the entity ID using a page-to-entity recommendation data structure generated by associating a set of canonical Web pages with the entity ID, associating a first set of relevant URLs to the set of canonical Web pages, associating a second set of relevant URLs to the first set of relevant URLs, and mapping the entity ID to the first set of relevant URLs and the second set of relevant URLs; and

(3) determining one or more search queries mapped to the entity ID;

generating a set of multi-domain recommendations related to the topic-of-interest, the set of multi-domain recommendations comprising at least a portion of the one or more entities, the one or more URLs, and the one or more search queries; and

presenting the set of multi-domain recommendations on the Web page being viewed by the user.

2. The media of claim 1, wherein the multi-domain recommendation data structure comprises a mapping between one or more entity IDs and relevant, related entities, URLs, and search queries.

3. The media of claim 1, wherein identifying the topic-of-interest on the Web page currently being viewed by the user includes identifying an author of the Web page.

4. The media of claim 3, wherein identifying the author of the Web page comprises:

extracting textual information associated with the Web page; and

determining that at least a portion of the textual information comprises the author'"'"'s name.

5. The media of claim 4, wherein determining the one or more entities mapped to the entity ID comprises determining other Web pages authored by the author.

6. The media of claim 1, wherein the entity ID represents at least one of a set of canonical Web pages or a Web page author.

7. The media of claim 6, wherein the set of canonical Web pages comprises one or more Web pages associated with a single domain name.

8. The media of claim 1, wherein presenting the set of multi-domain recommendations on the Web page currently being viewed by the user comprises presenting the set of multi-domain recommendations on a tool bar or a pop-up window.

9. The media of claim 1, wherein the set of multi-domain recommendations is presented upon receiving a user-initiated selection request.

10. The media of claim 1, wherein each recommendation within the set of multi-domain recommendations is associated with a selectable option.

11. The media of claim 10, further comprising:

receiving a selection of at least one recommendation of the set of multi-domain recommendations; and

navigating the user to one or more Web pages having content associated with the at least one recommendation of the set of multi-domain recommendations.

12. A computerized method carried out by a recommendation engine running on a processor for generating a multi-domain recommendation data structure, the method comprising:

accessing a data store to identify related uniform resource locators (URLs), wherein related URLs are identified by;

(1) identifying URLs frequently selected by users in response to a search query,

(2) identifying URLs frequently visited by users in a browser session, and

(3) identifying URLs linked to a source URL,

analyzing features of the identified URLs to determine a relevance of each of the identified URLs;

generating a page-to-page recommendation data structure by mapping each of the identified URLs to one or more of the related, relevant URLs; and

using the page-to-page recommendation data structure to generate a page-to-entity recommendation data structure, an entity-to-entity recommendation data structure, and a search query recommendation data structure, wherein using the page-to-page recommendation data structure to generate the page-to-entity recommendation data structure comprises;

(1) identifying a set of canonical Web pages associated with an entity ID;

(2) identifying a first set of relevant URLs mapped to the set of canonical Web pages using the page-to-page recommendation data structure;

(3) identifying a second set of relevant URLs mapped to the first set of relevant URLs using the page-to-page recommendation data structure; and

(4) generating the page-to-entity recommendation data structure by mapping the entity ID to the first set of relevant URLs and the second set of relevant URLs.

13. The method of claim 12, wherein the data store comprises query click logs, Web page hyperlink data, search log data, and entity ID data.

14. The method of claim 13, wherein the entity ID data comprises sets of canonical Web pages, each associated with an entity ID.

15. The method of claim 13, further comprising:

aggregating the page-to-page recommendation data structure, the page-to-entity recommendation data structure, the entity-to-entity recommendation data structure, and the search query recommendation data structure to generate the multi-domain recommendation data structure.

16. The method of claim 12, wherein analyzing features of the identified URLs to determine a relevance of the identified URLs comprises at least one of:

(1) determining a number of inlinks to each one of the identified URLs,

(2) determining a frequency visited for each one of the identified URLs, or

(3) determining a page rank of each one of the identified URLs.

17. The method of claim 12, wherein using the page-to-page recommendation data structure to generate the entity-to-entity recommendation data structure comprises:

identifying a first set of relevant URLs mapped to an entity ID using the page-to-entity recommendation data structure;

identifying a second set of relevant URLs mapped to the first set of URLs using the page-to-page recommendation data structure;

identifying one or more entity IDs mapped to the second set of relevant URLs using the page-to-entity recommendation data structure; and

generating the entity-to-entity recommendation data structure by mapping the entity ID to the one or more entity IDs.

18. The method of claim 17, wherein using the page-to-page recommendation data structure to generate the search query recommendation data structure comprises:

using the query click logs, identifying a first URL frequently selected in response to a search query;

using the page-to-page recommendation data structure, identifying relevant URLs mapped to the first URL;

using the query click logs, identifying search queries frequently associated with the relevant URLs;

using the page-to-entity recommendation data structure, identifying entity IDs mapped to the relevant URLs; and

generating the search query recommendation data structure by mapping the search queries to the entity IDs.

19. A system for generating and presenting multi-domain recommendations to a user viewing a topic-of-interest on a Web page, the system comprising:

a computing device associated with a recommendation engine having one or more processors and one or more computer-readable storage media; and

a data store coupled with the recommendation engine,

wherein the recommendation engine;

generates a page-to-page recommendation data structure, the page-to-page recommendation data structure comprising a mapping of at least a first uniform resource locator (URLs) to one or more related, relevant URLs;

generates a page-to-entity recommendation data structure, wherein generating the page-to-entity recommendation data structure comprises;

     1) identifying a set of canonical Web pages associated with an entity ID,

     2) identifying a first set of relevant URLs mapped to the set of canonical Web pages using the page-to-page recommendation data structure,

     3) identifying a second set of relevant URLs mapped to the first set of relevant URLs using the page-to-page recommendation data structure, and

     4) mapping the entity ID to the first set of relevant URLs and the second set of relevant URLs;

generates an entity-to-entity recommendation data structure, the entity-to-entity recommendation data structure comprising a mapping of the at least the first entity ID to one or more related entities;

generates a search query recommendation data structure, the search query recommendation data structure comprising a mapping of the at least the first entity ID to one or more related search queries;

aggregates the page-to-page recommendation data structure, the page-to-entity recommendation data structure, the entity-to-entity recommendation data structure, and the search query recommendation data structure to generate a multi-domain recommendation data structure, the multi-domain recommendation data structure comprising at least a mapping of the first entity ID to related, relevant entities, the first entity ID to related, relevant URLs, and the first entity ID to related, relevant search queries;

identifies the topic-of-interest on the Web page currently being viewed by the user;

maps the topic-of-interest to the first entity ID;

generates the multi-domain recommendations using the multi-domain recommendation data structure, the multi-domain recommendations comprising entities related to the topic-of-interest, URLs related to the topic-of-interest, and search queries related to the topic-of-interest; and

automatically presents the multi-domain recommendations on the Web page being viewed by the user.