Generating a user-specific ranking model on a user electronic device

• US 10,061,820 B2
• Filed: 01/27/2015
• Issued: 08/28/2018
• Est. Priority Date: 08/19/2014
• Status: Active Grant

• Current Assignee: YE Hub Armenia LLC
• Original Assignee: Yandex Europe AG (Yandex N.V.)
• Inventors: Logachev, Alexander Alexandrovich
• Primary Examiner(s): Beausoliel, Jr., Robert
• Assistant Examiner(s): Rayyan, Susan F
• Application Number:
  US15/100,337
• Publication Number:
  US 20160299899A1
• Time in Patent Office: 1,309 Days
• CPC Class Codes: G06F 16/24578 using ranking
• 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: G06N 20/00 Machine learning

Abstract

There is disclosed a method of generating a user-specific ranking model on an electronic device associated with a user. The method is executable on the electronic device. The method comprises: receiving, from a search engine server, via a communication network, an indication of a resource-specific feature; appreciating a user interaction with the web resource performed by the user using the electronic apparatus; based on the user interaction, determining a value parameter for the web resource; generating the user-specific ranking model on the basis of the value parameter and the resource-specific feature.

Claims

1. A method of generating a user-specific ranking model on an electronic device associated with a user, the method comprising:

receiving, by the electronic device, from a search engine server, via a communication network, an indication of a plurality of resource-specific features associated with a web resource, at least a portion of the plurality of the resource-specific features being only trackable at the electronic device, the search engine server using a generic ranking model;

receiving, by the electronic device, a user interaction with the web resource performed by the user using the electronic device;

based on the user interaction, determining, by the electronic device, a value parameter for the web resource;

generating, by a machine learning algorithm (MLA) on the electronic device, the user-specific ranking model on the basis of the value parameter and the plurality of resource-specific features;

transmitting to the search engine server the user-specific ranking model, the transmitting causing the search engine server to retrain the generic ranking model with the user-specific ranking model to obtain a combined ranking model;

sending, to the search engine server, a query from the user of the electronic device, causing the search engine server to;

retrieve search results associated with the query;

rank the search results by using the combined ranking model to generate a user-specific ranked search results list; and

transmit the user-specific ranked search results list to the electronic device.

2. The method of claim 1, wherein another portion of the plurality of resource-specific features are used by the generic ranking model used by the search engine server for ranking search results.

3. The method of claim 1, wherein said user interaction is indicative of a user preference associated with the web resource.

4. The method of claim 1, wherein said generating comprises applying the machine learning algorithm to create the user-specific ranking model using the plurality of resource-specific features and the value parameter as a training set.

5. The method of claim 1, wherein the at least portion of the plurality of resource-specific features are selected so that to optimize tracking of the user interaction on the electronic device.

6. The method of claim 1, wherein another portion of the plurality of resource-specific features are selected on the basis of the generic ranking model used by the search engine server to rank search results.

7. The method of claim 6, wherein said generic ranking model is in itself personalized based on a server-tracked user interaction associated with the another portion of the plurality of resource-specific features.

8. The method of claim 7, wherein said server-tracked user interaction associated with the another portion of the plurality of resource-specific features is of a different type than said user interaction tracked by the electronic device.

17. The method of claim 1, wherein:

the user interaction is a plurality of user interactions; and

the determining further comprises determining a plurality of value parameters based on the plurality of user interactions.

9. A method of processing a query by using a user-specific ranking model, the method executable at a search engine server, the search engine server being coupled to an electronic device via a communications network, the search engine server using a generic ranking model, the electronic device being associated with a user, the method comprising:

transmitting to the electronic device, via the communication network, an indication of a plurality of resource-specific features associated with a web resource, at least a portion of the plurality of the resource-specific features being only trackable at the electronic device;

receiving, from the electronic device, via the communication network, the user-specific ranking model, the user-specific ranking model having been generated by the electronic device by executing the steps of;

receiving, by the electronic device, a user interaction with the web resource performed by the user using the electronic device;

based on the user interaction, determining, by the electronic device, a value parameter for the web resource;

generating, by a machine learning algorithm (MLA) on the electronic device, the user-specific ranking model on the basis of the value parameter and the plurality of resource-specific features;

combining, by the search engine server, the user-specific ranking model with the generic ranking model to create a combined ranking model by retraining the generic ranking model using the user-specific ranking model;

receiving, by the search engine server, the query from the user of the electronic device;

retrieving, by the search engine server, search results associated with the query; and

ranking, by the search engine server, the search results by using the combined ranking model to generate a user-specific ranked search results list.

10. The method of claim 9, wherein another portion of the plurality of resource-specific features are used in a generic ranking model used by the search engine server for ranking search results.

11. The method of claim 9, wherein said selecting comprises choosing the at least the portion of the plurality of resource-specific features so that to optimize tracking of the user interaction on the electronic device.

12. The method of claim 9, wherein said user interaction is indicative of a user preference associated with the web resource.

13. The method of claim 12, further comprising:

sending, by the search engine server, the user-specific ranked search result list to the electronic device.

14. The method of claim 13, wherein said combining comprises:

adapting values of the generic ranking model corresponding to another portion of the plurality of resource-specific feature to correspond to the values used in the user-specific ranking model.

15. The method of claim 9, wherein:

the user interaction is a plurality of user interactions; and

the determining further comprises determining a plurality of value parameters based on the plurality of user interactions.

16. A server coupled to a communication network, the server comprising:

a communication interface for communication with an electronic device via the communication network,

a processor operationally connected with the communication interface,

the processor being configured to;

transmit to the electronic device, via the communication interface, an indication of a plurality of resource-specific features associated with a web resource, at least a portion of the plurality of the resource-specific features being only trackable at the electronic device, the search engine server using a generic ranking model;

receive from the communication device, via the communication interface, a user-specific ranking model, the user-specific ranking model having been generated by the electronic device by executing the steps of;

receiving, by the electronic device, a user interaction with the web resource performed by the user using the electronic device;

based on the user interaction, determining, by the electronic device, a value parameter for the web resource;

generating, by the electronic device, the user-specific ranking model on the basis of the value parameter and the plurality of resource-specific features;

combine the user-specific ranking model with the generic ranking model to create a combined ranking model by retraining the generic ranking model using the user-specific ranking model;

receive a query from the user of the electronic device;

retrieve search results associated with the query; and

rank the search results by using the combined ranking model to generate a user-specific ranked search results list.

18. The server of claim 16, wherein:

the user interaction is a plurality of user interactions; and

the determining further comprises determining a plurality of value parameters based on the plurality of user interactions.