Real-time online-learning object recommendation engine

Abstract

In one embodiment, a method includes receiving a request from a first user for a content page; in response to the request, identifying at least one content page, wherein the content page is associated with a page identifier; identifying a plurality of content items based at least in part on a plurality of content features associated with the content page; ranking the plurality of content items based at least in part on a plurality of user features associated with the first user; and delivering to the first user, with the requested content page, one or more of the plurality of content items as recommendations to the first user based on the ranking of the content items.

1 . A method comprising:
 by one or more computing devices, receiving a request from a first user for a content page;   by one or more computing devices, in response to the request, identifying at least one content page, wherein the content page is associated with a page identifier;   by one or more computing devices, identifying a plurality of content items based at least in part on a plurality of content features associated with the content page;   by one or more computing devices, ranking the plurality of content items based at least in part on a plurality of user features associated with the first user; and   by one or more computing devices, providing to the first user, with the requested content page, one or more of the plurality of content items as recommendations to the first user based on the ranking of the content items.   
     
     
         2 . The method of  claim 1 , wherein the content page is associated with a content context, the content context comprising a content page identifier. 
     
     
         3 . The method of  claim 1 , wherein the content page is associated with a content context, the content context comprising a list of one or more other users in the requested content. 
     
     
         4 . The method of  claim 1 , wherein the content page is associated with a content context, the content context comprising a list of one or more services for the content. 
     
     
         5 . The method of  claim 1 , wherein the plurality of user features comprise the first user's age. 
     
     
         6 . The method of  claim 1 , wherein the plurality of user features comprise the first user's gender. 
     
     
         7 . The method of  claim 1 , wherein the plurality of user features comprise the first user's last request of the content. 
     
     
         8 . The method of  claim 1 , wherein the content features comprise aggregate statistics for the content page. 
     
     
         9 . The method of  claim 8 , wherein the aggregate statistics comprise the click through rate for the content page. 
     
     
         10 . The method of  claim 8 , wherein the aggregate statistics comprise the conversion rate for one or more objects associated with the content page. 
     
     
         11 . The method of  claim 8 , wherein the aggregate statistics comprise the number of impressions in a predetermined period of time for one or more objects associated with the content page. 
     
     
         12 . The method of  claim 1 , wherein the ranking is performed by a recommendation model, the recommendation model comprising a set of weights for each of the user and content features calculated via one or more statistical models. 
     
     
         13 . The method of  claim 12 , wherein the one or more statistical models comprises logistic regression. 
     
     
         14 . The method of  claim 1 , further comprising:
 receiving an example comprising a positive signal or a negative signal for a particular set of user features and content features; and   updating the recommendation model based on the example.   
     
     
         15 . The method of  claim 14 , wherein:
 the example is positive if a viewing user having the particular set of user features fails to convert on a viewed content item having the particular set of content features after an impression, and   the example is negative if the viewing user having the particular set of user features converts on the viewed content item having the particular set of content features after the impression.   
     
     
         16 . The method of  claim 15 , wherein converting on the viewed object comprises clicking on a link displayed in a landing page of the viewed object. 
     
     
         17 . A non-transitory, computer-readable media comprising instructions operable, when executed by one or more computing systems, to:
 receive a request from a first user for a content page;   in response to the request, identify at least one content page, wherein the content page is associated with a page identifier;   identify a plurality of content items based at least in part on a plurality of content features associated with the content page;   rank the plurality of content items based at least in part on a plurality of user features associated with the first user; and   provide to the first user, with the requested content page, one or more of the plurality of content items as recommendations to the first user based on the ranking of the content items.   
     
     
         18 . The media of  claim 17 , wherein the content page is associated with a content context, the content context comprising a content page identifier. 
     
     
         19 . The media of  claim 17 , the instructions further operable, when executed by one or more computing systems, to:
 receive an example comprising a positive signal or a negative signal for a particular set of user features and content features; and   update the recommendation model based on the example.   
     
     
         20 . A system comprising:
 one or more processors; and   one or more computer-readable non-transitory storage media coupled to one or more of the processors and comprising instructions operable when executed by one or more of the processors to cause the system to:
 receive a request from a first user for a content page; 
 in response to the request, identify at least one content page, wherein the content page is associated with a page identifier; 
 identify a plurality of content items based at least in part on a plurality of content features associated with the content page; 
 rank the plurality of content items based at least in part on a plurality of user features associated with the first user; and 
 provide to the first user, with the requested content page, one or more of the plurality of content items as recommendations to the first user based on the ranking of the content items.