Client-server hybrid AI scores for customized actions

• US 10,373,046 B2
• Filed: 01/06/2017
• Issued: 08/06/2019
• Est. Priority Date: 01/06/2017
• Status: Active Grant

• Current Assignee: Salesforce.com, Inc.
• Original Assignee: Salesforce.com, Inc.
• Inventors: Vijayant, Amit, Jagota, Arun Kumar
• Primary Examiner(s): Lee, Gil H.
• Application Number:
  US15/400,331
• Publication Number:
  US 20180198889A1
• Time in Patent Office: 942 Days
• CPC Class Codes: G06N 20/00 Machine learning
• CPC Class Codes: G06N 3/004 Artificial life, i.e. computing arrangements simulating life
• CPC Class Codes: H04L 67/02 based on web technology, e.g. hypertext transfer protocol [HTTP]
• CPC Class Codes: H04L 67/306 User profiles
• CPC Class Codes: H04L 67/535 Tracking the activity of the user network monitoring arrangements H04L43/00; recording of computer activity G06F11/34

Abstract

Client-server hybrid A.I. scores for customized actions are described. A client generates client scores corresponding to client customized actions by applying a user-specific model to an action received from a user, the user-specific model based on at least one historical action received from the user. The client requests a server to provide server scores corresponding to server customized actions by applying a cross-user model to the action received from the user, the cross-user model based on historical actions associated with server users. The client generates hybrid scores corresponding to hybrid customized actions by combining the client scores with the server scores, in response to receiving the server scores from the server. The client causes the hybrid customized actions to be outputted based on the corresponding hybrid scores.

Claims

1. A system comprising:

one or more processors; and

a non-transitory computer readable medium storing a plurality of instructions, which when executed, cause the one or more processors to;

generate, by a client computing system, a plurality of client scores corresponding to a plurality of client customized actions by applying a user-specific model residing on the client computing system to an action received from a user, the user-specific model being based on at least one historical action received from the user, the user-specific model including modifications based upon a cross-user model received from a server, the cross-user model being based on a plurality of historical actions associated with a plurality of server users and including changes in the cross-user model that establish a new trend detected by the server prior to sending the cross-user model to the client computing system;

request, by the client computing system, the server to provide a plurality of server scores corresponding to a plurality of server customized actions by applying the cross-user model to the action received from the user;

combine, by the client computing system, the plurality of client scores with the plurality of server scores received from the server to generate a plurality of hybrid scores corresponding to a plurality of hybrid customized actions, the plurality of hybrid scores being different than the pluralities of client and server scores; and

cause, by the client computing system, the plurality of hybrid customized actions to be outputted based on the plurality of corresponding hybrid scores.

2. The system of claim 1, wherein requesting the server to provide the plurality of server scores comprises providing the server with data associated with the user.

3. The system of claim 1, wherein the plurality of server scores are further based on applying a content model to the action from the user, the content model based on data associated with server content.

4. The system of claim 1, wherein combining the plurality of client scores with the plurality of server scores is based on a client weight assigned to the plurality of client scores and a server weight assigned to the plurality of server scores, the client weight being different from the server weight, and the client weight being based on at least one of a count of any historical user actions associated with the action and a temporal proximity of any historical user action associated with the action.

5. The system of claim 4, wherein the client weight is further based on an association between a historical hybrid customized action and a historical user action subsequent to the historical hybrid customized action.

6. The system of claim 1, comprising further instructions, which when executed, cause the one or more processors to send, by the client, the user-specific model to the server, thereby enabling the server to modify the cross-user model based on the user-specific model.

7. A computer program product comprising a non-transitory computer readable medium storing computer-readable program code to be executed by one or more processors, the program code including instructions to:

generate, by a client computing system, a plurality of client scores corresponding to a plurality of client customized actions by applying a user-specific model residing on the client computing system to an action received from a user, the user-specific model being based on at least one historical action received from the user, the user-specific model including modifications based upon a cross-user model received from a server, the cross-user model being based on a plurality of historical actions associated with a plurality of server users and including changes in the cross-user model that establish a new trend detected by the server prior to sending the cross-user model to the client computing system;

request, by the client computing system, the server to provide a plurality of server scores corresponding to a plurality of server customized actions by applying the cross-user model to the action received from the user;

combine, by the client computing system, the plurality of client scores with the plurality of server scores received from the server to generate a plurality of hybrid scores corresponding to a plurality of hybrid customized actions, the plurality of hybrid scores being different than the pluralities of client and server scores; and

cause, by the client computing system, the plurality of hybrid customized actions to be outputted based on the plurality of corresponding hybrid scores.

8. The computer program product of claim 7, wherein requesting the server to provide the plurality of server scores comprises providing the server with data associated with the user.

9. The computer program product of claim 7, wherein the plurality of server scores are further based on applying a content model to the action from the user, the content model based on data associated with server content.

10. The computer program product of claim 7, wherein combining the plurality of client scores with the plurality of server scores is based on a client weight assigned to the plurality of client scores and a server weight assigned to the plurality of server scores, the client weight being different from the server weight, and the client weight being based on at least one of a count of any historical user actions associated with the action and a temporal proximity of any historical user action associated with the action.

11. The computer program product of claim 10, wherein the client weight is further based on an association between a historical hybrid customized action and a historical user action subsequent to the historical hybrid customized action.

12. The computer program product of claim 7, wherein the program code comprises further instructions to send, by the client, the user-specific model to the server, thereby enabling the server to modify the cross-user model based on the user-specific model.

13. A method comprising:

generating, by a client computing system, a plurality of client scores corresponding to a plurality of client customized actions by applying a user-specific model residing on the client computing system to an action received from a user, the user-specific model being based on at least one historical action received from the user, the user-specific model including modifications based upon a cross-user model received from a server, the cross-user model being based on a plurality of historical actions associated with a plurality of server users and including changes in the cross-user model that establish a new trend detected by the server prior to sending the cross-user model to the client computing system;

requesting, by the client computing system, the server to provide a plurality of server scores corresponding to a plurality of server customized actions by applying the cross-user model to the action received from the user;

combining, by the client computing system, the plurality of client scores with the plurality of server scores received from the server to generate a plurality of hybrid scores corresponding to a plurality of hybrid customized actions, the plurality of hybrid scores being different than the pluralities of client and server scores; and

causing, by the client computing system, the plurality of hybrid customized actions to be outputted based on the plurality of corresponding hybrid scores.

14. The method of claim 13, wherein requesting the server to provide the plurality of server scores comprises providing the server with data associated with the user.

15. The method of claim 13, wherein the plurality of server scores are further based on applying a content model to the action from the user, the content model based on data associated with server content.

16. The method of claim 13, wherein combining the plurality of client scores with the plurality of server scores comprises is based on a client weight assigned to the plurality of client scores and a server weight assigned to the plurality of server scores, the client weight being different from the server weight, the client weight being based on at least one of a count of any historical user actions associated with the action, a temporal proximity of any historical user action associated with the action, and an association between a historical hybrid customized action and a historical user action subsequent to the historical hybrid customized action.

17. The method of claim 13, wherein the method further comprises sending, by the client, the user-specific model to the server, thereby enabling the server to modify the cross-user model based on the user-specific model.