News feed ranking model based on social information of viewer

Patent metadata

Field	Value
Publication number	US-9582786-B2
Application number	US-201113194773-A
Country	US
Kind code	B2
Filing date	Jul 29, 2011
Priority date	Jul 29, 2011
Publication date	Feb 28, 2017
Grant date	Feb 28, 2017

Abstract

Machine learning models are used for ranking news feed stories presented to users of a social networking system. The social networking system divides its users into different sets, for example, based on demographic characteristics of the users and generates one model for each set of users. The models are periodically retrained. The news feed ranking model may rank news feeds for a user based on information describing other users connected to the user in the social networking system. Information describing other users connected to the user includes interactions of the other users with objects associated with news feed stories. These interactions include commenting on a news feed story, liking a news feed story, or retrieving information, for example, images, videos associated with a news feed story.

First claim

The invention claimed is:

1. A computer-implemented method comprising:
receiving information describing interactions of users of a social networking system with a plurality of news feed stories maintained by the social networking system;
identifying one or more news feed stories from the plurality of news feed stories for presentation to a viewer;
for each of the one or more identified news feed stories, identifying a first set of features describing interactions of other users of the social networking system with whom the viewer has previously established a connection, the features from the first set comprising a first aggregate measure based on interactions of the other users, wherein the interactions of the other users are with objects described in the news feed story, the first aggregate measure weighing interactions of the other users based on their affinity with the viewer;
training a first news feed ranking model using the first set of features, the first news feed ranking model configured to rank candidate news feed stories selected for presentation to the viewer;
for each of the one or more identified news feed stories, identifying a second set of features describing interactions of the viewer, the features from the second set comprising a second aggregate measure based on interactions of the viewer, wherein the interactions of the viewer are with objects described in the news feed story;
training a second news feed ranking model using the second set of features, the second news feed ranking model configured to rank candidate news feed stories selected for presentation to the viewer; and
determining whether to use the first news feed ranking model or the second news feed ranking model to rank candidate news feed stories selected for presentation to the viewer based on a number of interactions of the viewer with the objects described in the news feed story being below a predetermined threshold.

2. The computer-implemented method of claim 1 , wherein the first set of features further describe interactions of the other users with other news feed stories similar to the news feed story.

3. The computer-implemented method of claim 1 , wherein the second set of features further describe interactions of the viewer with other news feed stories similar to the news feed story.

4. The computer-implemented method of claim 1 , wherein the first set of features further describes demographic information of the other users.

5. The computer-implemented method of claim 1 , wherein the interactions of the users of the social networking system comprise positive interactions of the users with a first set of news feed stories of the plurality of news feed stories and negative interactions of the users with a second set of news feed stories of the plurality of news feed stories, the positive interactions indicative of user interest in the first set of news feed stories, and the negative interactions indicative of lack of user interest in the second set of news feed stories.

6. The computer-implemented method of claim 5 , wherein at least one of the positive interactions is retrieving additional information describing one of the news feed stories of the first set, recommending one of the news feed stories of the first set, liking one of the news feed stories of the first set, commenting on one of the new feed stories of the first set, or selecting a link from one of the news feed stories of the first set.

7. The computer-implemented method of claim 5 , wherein at least one of the negative interactions is deleting one of the news feed stories of the second set from a profile page of one of the users of the social networking system, hiding one of the news feed stories of the second set, or ignoring one of the news feed stories of the second set.

8. The computer-implemented method of claim 1 , wherein the received information further describes interactions of the users of the social networking system with one or more objects, each of the one or more objects associated with at least one of the news feed stories, and wherein for a particular news feed story, the first set of features and the second set of features are based at least in part on an object representing an entity described in the particular news feed story.

9. A computer-implemented method comprising:
selecting a set of candidate news feed stories maintained by a social networking system for presentation to a viewer, each candidate news feed story describing an interaction of a user with one or more objects;
for each candidate news feed story, identifying a first set of features describing interactions of other users of the social networking system with whom the viewer has previously established a connection, the features from the first set comprising a first aggregate measure based on interactions of the other users with objects described in the candidate news feed story, the first aggregate measure weighing interactions of the other users based on their affinity with the viewer;
providing the first set of features as input to a first news feed ranking model trained to rank news feed stories for presentation to the viewer;
for each candidate news feed story, identifying a second set of features describing interactions of the viewer, the features from the second set comprising a second aggregate measure based on interactions of the viewer, wherein the interactions of the viewer are with objects described in the candidate news feed story;
providing the second set of features as input to a second news feed ranking model trained to rank news feed stories for presentation to the viewer; and
determining whether to use the first news feed ranking model or the second news feed ranking model to rank the set of candidate news feed stories based on a number of interactions of the viewer with objects described in the candidate news feed story being below a predetermined threshold.

10. The computer-implemented method of claim 9 , wherein the first set of features further describe interactions of the other users with other news feed stories similar to the news feed story.

11. The computer-implemented method of claim 9 , wherein the second set of features further describe interactions of the viewer with other news feed stories similar to the news feed story.

12. The computer-implemented method of claim 9 , wherein the first set of features describes demographic information of the other users.

13. The computer-implemented method of claim 9 , wherein the interactions of the other users comprise at least one of retrieving additional information describing a candidate news feed story, recommending a candidate news feed story, liking a candidate news feed story, commenting on a candidate news feed story, selecting a link from a candidate news feed story, deleting a candidate news feed story from a profile page of a user of the social networking system, hiding a candidate news feed story, or ignoring a candidate news feed story.

14. The computer-implemented method of claim 9 , wherein the interactions of the other users of the social networking system comprise positive interactions of the users with a first set of news feed stories of the plurality of news feed stories and negative interactions of the users with a second set of news feed stories of the plurality of news feed stories, the positive interactions indicative of user interest in the first set of news feed stories, and the negative interactions indicative of lack of user interest in the second set of news feed stories.

15. The computer-implemented method of claim 14 , wherein at least one of the positive interactions is retrieving additi

Assignees

• Gubin Max
• Kao Wayne
• Vickrey David
• Maykov Alexey
• Facebook Inc

Inventors

• Gubin Max
• Kao Wayne
• Vickrey David
• Maykov Alexey

Classifications

• G06Q10/40

  Business processes related to social networking or social networking services · CPC title

• G06F17/30867

  Physics · mapped topic

• G06F11/34

  Recording or statistical evaluation of computer activity, e.g. of down time, of input/output operation {; Recording or statistical evaluation of user activity, e.g. usability assessment} · CPC title

• H04L67/306

  User profiles · CPC title

• G06Q10/10 Primary

  Office automation; Time management · CPC title