Using combined coefficients for viral action optimization in an on-line social network

Patent metadata

Field	Value
Publication number	US-10324937-B2
Application number	US-201615221195-A
Country	US
Kind code	B2
Filing date	Jul 27, 2016
Priority date	Feb 10, 2016
Publication date	Jun 18, 2019
Grant date	Jun 18, 2019

Abstract

A news feed system provided with an on-line social network system determines that a news feed is to be constructed for a viewer. The news feed system accesses the viewer's profile and other information associated with the viewer, accesses an inventory of activities that have been identified as potentially of interest to the viewer, and calculates relevance score for each item inventory of activities using the combined coefficients methodology. The activities are then arranged for presentation to the viewer via a news feed web page, using respective calculated relevance scores.

First claim

The invention claimed is:

1. A computer-implemented method comprising:
learning a first logistic regression model to obtain a first coefficient vector, the first coefficient vector being with respect to probability of a viewer interaction with an item presented to the viewer via a news feed web page;
learning a second logistic regression to obtain a second coefficient vector, the second coefficient vector being with respect to probability of a viewer performing a viral action on an item presented to the viewer via a news feed web page;
executing simulations each time varying a value of a tradeoff parameter that indicates importance of viral actions as compared to interactions;
constructing a tradeoff curve based on the result of the simulations;
selecting a value of the tradeoff parameter to be used in the calculating of the combined coefficient based on information indicated by the tradeoff curve;
calculating a combined coefficient based on the first coefficient vector, the second coefficient vector, and the tradeoff parameter;
using at least one processor, executing a personalization model to obtain a respective relevance rank for each item in an inventory of activities identified as potentially of interest to a focus viewer, using the combined coefficient and respective feature vectors, a feature vector is constructed with respect to the focus viewer and a given item from the inventory of activities using signals characterizing the focus viewer and the given item; and
ordering the items in the inventory for presentation to the focus viewer based on the respective relevance ranks of the items in the inventory.

2. The method of claim 1 , comprising constructing a news feed web page that includes items in the inventory ordered based on the respective relevance ranks.

3. The method of claim 1 , comprising causing presentation of the news feed web page on a display device of the focus viewer.

4. The method of claim 1 , wherein the learning of the first logistic regression model to obtain the first coefficient vector is using historical data obtained in the on-line social network system.

5. The method of claim 1 , wherein the learning of the second logistic regression model to obtain the second coefficient vector is using historical data obtained in the on-line social network system.

6. The method of claim 1 , wherein a viral action is an action that results in an additional item being included in an inventory of updates for another member in the on-line social network system.

7. The method of claim 1 , wherein an interaction is any action by a viewer with respect to an item presented to the viewer via a news feed web page.

8. The method of claim 1 , wherein the executing of simulation comprises utilizing a random session where a viewer is presented with a feed that includes randomly chosen activities.

9. A computer-implemented system comprising:
one or more processors; and
a non-transitory computer readable storage medium comprising instructions that when executed by the one or processors cause the one or more processors to perform operations comprising:
learning a first logistic regression model to obtain a first coefficient vector, the first coefficient vector being with respect to probability of a viewer interaction with an item presented to the viewer via a news feed web page;
learning a second logistic regression to obtain a second coefficient vector, the second coefficient vector being with respect to probability of a viewer performing a viral action on an item presented to the viewer via a news feed web page;
executing simulations each time varying a value of a tradeoff parameter that indicates importance of viral actions as compared to interactions;
constructing a tradeoff curve based on the result of the simulations;
selecting a value of the tradeoff parameter to be used in the calculating of the combined coefficient based on information indicated by the tradeoff curve; and
calculating a combined coefficient based on the first coefficient vector, the second coefficient vector, and the tradeoff parameter;
executing a personalization model to obtain a respective relevance rank for each item in an inventory of activities identified as potentially of interest to a focus viewer, using the combined coefficient and respective feature vectors, a feature vector is constructed with respect to the focus viewer and a given item from the inventory of activities using signals characterizing the focus viewer and the given item; and
ordering the items in the inventory for presentation to the focus viewer based on the respective relevance ranks of the items in the inventory.

10. The system of claim 9 , comprising constructing a news feed web page that includes items in the inventory ordered based on the respective relevance ranks.

11. The system of claim 9 , comprising causing presentation of the news feed web page on a display device of the focus viewer.

12. The system of claim 9 , wherein the learning of the first logistic regression model to obtain the first coefficient vector is using historical data obtained in the on-line social network system.

13. The system of claim 9 , wherein the learning of the second logistic regression model to obtain the second coefficient vector is using historical data obtained in the on-line social network system.

14. The system of claim 9 , wherein a viral action is an action that results in an additional item being included in an inventory of updates for another member in the on-line social network system.

15. The system of claim 9 , wherein an interaction is any action by a viewer with respect to an item presented to the viewer via a news feed web page.

16. The system of claim 9 , wherein the executing of simulation comprises utilizing a random session where a viewer is presented with a feed that includes randomly chosen activities.

17. A machine-readable non-transitory storage medium having instruction data executable by a machine to cause the machine to perform operations comprising:
learning a first logistic regression model to obtain a first coefficient vector, the first coefficient vector being with respect to probability of a viewer interaction with an item presented to the viewer via a news feed web page;
learning a second logistic regression to obtain a second coefficient vector, the second coefficient vector being with respect to probability of a viewer performing a viral action on an item presented to the viewer via a news feed web page;
executing simulations each time varying a value of a tradeoff parameter that indicates importance of viral actions as compared to interactions;
constructing a tradeoff curve based on the result of the simulations;
selecting a value of the tradeoff parameter to be used in the calculating of the combined coefficient based on information indicated by the tradeoff curve;
calculating a combined coefficient based on the first coefficient vector, the second coefficient vector, and the tradeoff parameter;
executing a personalization model to obtain a respective relevance rank for each item in an inventory of activities identified as potentially of interest to a focus viewer, using the combined coefficient and respective feature vectors, a feature vector is constructed with respect to the focus viewer and a given item from the inventory of activities using signals characterizing the focus viewer and the given item; and
ordering the items in the inventory for presentation to the focus viewer based on the respective relevance ranks of the items in the inventory.

Assignees

• Microsoft Technology Licensing Llc

Inventors

• Yang Jaewon
• Tang Liang
• Chen Bee-Chung

Classifications

• G06N7/01

  Probabilistic graphical models, e.g. probabilistic networks · CPC title

• G06N20/00

  Machine learning · CPC title

• G06F16/9535

  Search customisation based on user profiles and personalisation · CPC title

• G06F16/248

  Presentation of query results · CPC title

• H04L67/306 Primary

  User profiles · CPC title