Context sensitive ranking

• US 8,751,470 B1
• Filed: 05/31/2012
• Issued: 06/10/2014
• Est. Priority Date: 04/02/2008
• Status: Active Grant

• Current Assignee: Google LLC (Alphabet Inc.)
• Original Assignee: Google Inc. (Alphabet Inc.)
• Inventors: Garg, Ashutosh, Dhamdhere, Kedar
• Primary Examiner(s): Bhatia, Ajay
• Assistant Examiner(s): Le, Jessica N
• Application Number:
  US13/485,150
• Time in Patent Office: 740 Days
• US Class Current: 707/705
• CPC Class Codes: G06F 16/10 File systems; File servers
• CPC Class Codes: G06F 16/332 Query formulation
• CPC Class Codes: G06F 16/3325 Reformulation based on results of preceding query
• CPC Class Codes: G06F 16/3334 Selection or weighting of terms from queries, including natural language queries
• CPC Class Codes: G06F 16/3349 Reuse of stored results of previous queries
• CPC Class Codes: G06F 16/93 Document management systems
• 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: G06F 16/9538 Presentation of query results
• CPC Class Codes: G06Q 30/02 Marketing; Price estimation or determination; Fundraising

Abstract

Methods, systems, and apparatus, including computer program products, in which context can be used to rank search results. Context associated with a user session can be identified. A search query received during the user session can be used to identify a contextual click model based upon the context associated with the user session.

Claims

1. A computer-implemented method, comprising:

determining, for a current user session and by one or more computers, a first context that is indicative of a category of user interest for the current user session, the determination being based on content provided to a particular user during the current user session, wherein determining the first context comprises;

identifying previous user sessions of one or more other users, each previous user session including a selection of a hub page, the hub page being a page that has been identified as part of a particular context based at least in part on characteristics of the hub page;

designating a plurality of web pages as part of the particular context, the designation being based upon each of the plurality of web pages having been requested by one or more of the other users during the previous user sessions that included selection of the hub page;

determining that the content provided to the particular user during the current user session includes one of the plurality of web pages that were designated as part of the particular context; and

in response to determining that the content includes one of the plurality of web pages, designating the particular context to be the first context of the current user session;

receiving, during the current user session and after designating the particular context to be the first context, a user search query from the particular user; and

providing, by the one or more computers and in response to receiving the user search query, ranked search results for the user search query, the ranking being based at least in part on a first contextual click model for the first context.

2. The computer-implemented method of claim 1, further comprising identifying the first contextual click model, the identification comprising:

identifying previous search sessions that include selection of a web page that is identified as part of the first context;

identifying occurrences of the user search query in the identified previous search sessions;

identifying previously selected web pages that were selected in previous sessions as a result of the user search query; and

wherein the first contextual click model comprises the previously selected web pages and a selection metric defining a rate at which the previously selected web pages were selected as a result of the user search query.

3. The computer-implemented method of claim 2, wherein the first contextual click model further comprises a click quality metric for the selections of each of the previously selected web pages.

4. The computer-implemented method of claim 1, further comprising:

determining, during the current user session and by the one or more computers, a second context for the current user session by comparing web pages included in the content provided to the particular user during the current user session with a plurality of web pages defining the second context;

identifying, by the one or more computers, a second contextual click model for the user search query based upon content provided to past users during sessions that were previously identified as being associated with the second context; and

merging the first contextual click model and the second contextual click model.

5. The computer-implemented method of claim 4, wherein merging the first contextual click model and the second contextual click model comprises scaling a selection metric for the first contextual click model based upon a probability that the first contextual click model applies and scaling a selection metric for the second contextual click model based upon a probability that the second contextual click model applies.

6. The computer-implemented method of claim 1, further comprising generating a plurality of contextual click models for respective search queries associated with the first context, each contextual click model being operable to provide weightings for search results for respective search queries, and wherein the first contextual click model is one of the plurality of contextual click models.

7. The computer-implemented method of claim 1, wherein the first context comprises one or more of a news context, a shopping context, a travel context or an educational context, wherein the news context is determined when it is identified that the user is navigating sites related to news content, the shopping context is determined when it is identified that the user is navigating sites related to shopping content, the travel context is determined when it is identified that the user is navigating travel content and the educational context is determined when it is identified that the user is navigating educational content;

and

wherein the first contextual click model associated with the first context is operable to increase the rankings of search results referencing content related to the first context.

8. The computer-implemented method of claim 1, further comprising:

identifying search results responsive to the user search query; and

ranking the search results according to the first contextual click model.

9. The computer-implemented method of claim 1, wherein the characteristics of the hub page comprise an explicit context identifier provided by one or more of:

a hub page publisher; and

a context sensitive ranking system administrator.

10. A system, comprising:

a data processing apparatus; and

a data store storing instructions that, when executed by the data processing apparatus, cause the data processing apparatus to perform operations comprising;

determining, for a current user session, a first context that is indicative of a category of user interest for the current user session, the determination being based on content provided to a particular user during the current user session, wherein determining the first context comprises;

identifying previous user sessions of one or more other users, each previous user session including a selection of a hub page, the hub page being a page that has been identified as part of a particular context based at least in part on characteristics of the hub page;

designating a plurality of web pages as part of the particular context, the designation being based upon each of the plurality of web pages having been requested by one or more of the other users during the previous user sessions that included selection of the hub page;

determining that the content provided to the particular user during the current user session includes one of the plurality of web pages that were designated as part of the particular context; and

in response to determining that the content includes one of the plurality of web pages, designating the particular context to be the first context of the current user session;

receiving, during the current user session and after designating the particular context to be the first context, a user search query from the particular user; and

providing, in response to receiving the user search query, ranked search results for the user search query, the ranking being based at least in part on a first contextual click model for the first context.

11. The system of claim 10, the operations further comprising identifying the first contextual click model, the identification comprising:

identifying previous search sessions that include selection of a web page that is identified as part of the first context;

identifying occurrences of the user search query in the identified previous search sessions;

identifying previously selected web pages that were selected in previous sessions as a result of the user search query; and

wherein the first contextual click model comprises the previously selected web pages and a selection metric defining a rate at which the previously selected web pages were selected as a result of the user search query.

12. The system of claim 11, wherein the first contextual click model further comprises a click quality metric for the selections of each web page.

13. The system of claim 10, the operations further comprising:

determining, during the current user session, a second context for the current user session by comparing web pages included in the content provided to the particular user during the current user session with a plurality of web pages defining the second context;

identifying a second contextual click model for the user search query based upon content provided to past users during sessions that were previously identified as being associated with the second context; and

merging the first contextual click model and the second contextual click model.

14. The system of claim 13, wherein merging the first contextual click model and the second contextual click model comprises scaling a selection metric for the first contextual click model based upon a probability that the first contextual click model applies and scaling a selection metric for the second contextual click model based upon a probability that the second contextual click model applies.

15. The system of claim 10, the operations further comprising generating a plurality of contextual click models for respective search queries associated with the first context, each contextual click model being operable to provide weightings for search results for respective search queries, and wherein the first contextual click model is one of the plurality of contextual click models.

16. The system of claim 10, wherein the first context comprises one or more of a news context, a shopping context, a travel context or an educational context, wherein the news context is determined when it is identified that the user is navigating sites related to news content, the shopping context is determined when it is identified that the user is navigating sites related to shopping content, the travel context is determined when it is identified that the user is navigating travel content and the educational context is determined when it is identified that the user is navigating educational content;

and

wherein the first contextual click model associated with the first context is operable to increase the rankings of search results referencing content related to the first context.

17. The system of claim 10, wherein the current user session is defined by a predetermined maximum period of time between the user search query and a second search query.

18. The system of claim 10, wherein the current user session is defined by a login and logout time associated with the particular user.

19. The system of claim 10, the operations further comprising:

generating a plurality of contextual click models for respective search queries associated with the first context, each contextual click model being operable to provide weightings of search results for respective search queries.

20. The system of claim 10, the operations further comprising:

identifying search results responsive to the user search query; and

ranking the search results according to the first contextual click model.

21. The system of claim 10, wherein the characteristics of the hub page comprise an explicit context identifier provided by one or more of:

a hub page publisher; and

a context sensitive ranking system administrator.

22. One or more non-transitory computer readable media, comprising instructions that when executed by data processing apparatus cause the data processing apparatus to perform operations comprising:

determining, for a current user session, a first context that is indicative of a category of user interest for the current user session, the determination being based on content provided to a particular user during the current user session, wherein determining the first context comprises;

identifying previous user sessions of one or more other users, each previous user session including a selection of a hub page, the hub page being a page that has been identified as part of a particular context based at least in part on characteristics of the hub page;

designating a plurality of web pages as part of the particular context, the designation being based upon each of the plurality of web pages having been requested by one or more of the other users during the previous user sessions that included selection of the hub page;

determining that the content provided to the particular user during the current user session includes one of the plurality of web pages that were designated as part of the particular context; and

in response to determining that the content includes one of the plurality of web pages, designating the particular context to be the first context of the current user session;

receiving, during the current user session and after designating the particular context to be the first context, a user search query from the particular user; and

providing, in response to receiving the user search query, ranked search results for the user search query, the ranking being based at least in part on a first contextual click model for the first context.

23. A computer-implemented method, comprising:

determining, for a user session and by one or more computers, a first context that is indicative of a category of user interest for the user session, the determination being based on content provided to a particular user during the user session, wherein determining the first context comprises;

identifying a plurality of web pages as part of a particular context, each of the web pages being identified based upon inclusion of the web page in one or more of a plurality of click models associated with the particular context, wherein each click model is derived from previous search sessions of one or more other users;

determining that the content provided to the particular user during the user session comprises one of the plurality of web pages that were identified as part of the particular context; and

in response to determining that the content comprises one of the plurality of web pages, designating the particular context to be the first context of the user session;

receiving, during the user session and after designating the first context of the user session, a first search query from the particular user;

providing, by the one or more computers and in response to receiving the first search query, a first contextual click model associated with the first context to a search engine, wherein search results for the first search query are ranked based at least in part on the first contextual click model.

24. The method of claim 23, wherein the user session is defined by a predetermined maximum period of time between the first search query and a second search query.

25. The method of claim 23, wherein the user session is defined by a login and logout time associated with the particular user.

26. The method of claim 23, further comprising:

generating a plurality of contextual click models for respective search queries associated with the first context, each contextual click model being operable to provide weightings of search results for respective search queries.

27. The method of claim 23, further comprising:

determining that a second search query was not received within a predetermined period of time of the first search query; and

bypassing the provision of the first contextual click model.

28. A system, comprising:

a data processing apparatus; and

a data store storing instructions that, when executed by the data processing apparatus, cause the data processing apparatus to perform operations comprising;

determining, for a user session, a first context that is indicative of a category of user interest for the user session, the determination being based on content provided to a particular user during the user session, wherein determining the first context comprises;

identifying a plurality of web pages as part of a particular context, each of the web pages being identified based upon inclusion of the web page in one or more of a plurality of click models associated with the particular context, wherein each click model is derived from previous search sessions of one or more other users;

determining that the content provided to the particular user during the user session comprises one of the plurality of web pages that were identified as part of the particular context; and

in response to determining that the content comprises one of the plurality of web pages, designating the particular context to be the first context of the user session;

receiving, during the user session and after designating the first context of the user session, a first search query from the particular user;

providing, in response to receiving the first search query, a first contextual click model associated with the first context to a search engine, wherein search results for the first search query are ranked based at least in part on the first contextual click model.

29. The system of claim 28, wherein the user session is defined by a predetermined maximum period of time between the first search query and a second search query.

30. The system of claim 28, wherein the user session is defined by a login and logout time associated with the particular user.

31. The system of claim 28, the operations further comprising:

generating a plurality of contextual click models for respective search queries associated with the first context, each contextual click model being operable to provide weightings of search results for respective search queries.

32. The system of claim 28, the operations further comprising:

determining that a second search query was not received within a predetermined period of time of the first search query; and

bypassing the provision of the first contextual click model.

33. One or more non-transitory computer readable media, comprising instructions that when executed by data processing apparatus cause the data processing apparatus to perform operations comprising:

determining, for a user session, a first context that is indicative of a category of user interest for the user session, the determination being based on content provided to a particular user during the user session, wherein determining the first context comprises;

identifying a plurality of web pages as part of a particular context, each of the web pages being identified based upon inclusion of the web page in one or more of a plurality of click models associated with the particular context, wherein each click model is derived from previous search sessions of one or more other users;

determining that the content provided to the particular user during the user session comprises one of the plurality of web pages that were identified as part of the particular context; and

in response to determining that the content comprises one of the plurality of web pages, designating the particular context to be the first context of the user session;

receiving, during the user session and after designating the first context of the user session, a first search query from the particular user;

providing, in response to receiving the first search query, a first contextual click model associated with the first context to a search engine, wherein search results for the first search query are ranked based at least in part on the first contextual click model.