Ranking content for user engagement

Patent metadata

Field	Value
Publication number	US-10296642-B1
Application number	US-201514732356-A
Country	US
Kind code	B1
Filing date	Jun 5, 2015
Priority date	Jun 5, 2015
Publication date	May 21, 2019
Grant date	May 21, 2019

Abstract

A system comprising a processor and a memory storing instructions that, when executed, cause the system to receive a record of data describing user engagement with content items in an online service; prepare the record of data for generating a model characterizing a network effect of a user interaction with a content item in the online service; generate the model characterizing the network effect of the user interaction with the content item in the online service; generate a predicted network effect score for a plurality of content items based on the model; organize the plurality of content items based on the predicted network effect score; and transmit the plurality of organized content items for presentation to a user. The disclosure also includes similar methods and computer program products.

First claim

What is claimed is:

1. A computer-implemented method comprising:
receiving, using one or more processors, a record of data describing user engagement with a plurality of content items in an online service;
identifying, from the record of data, a first user interaction with a content item and a sequence of interactions resulted from the first user interaction in the online service;
generating a graph of interactions with the content item in the online service, wherein a vertex of the graph represents a user interaction of identified interactions and an edge represents a causal relation between the identified interactions;
calculating an engagement score for an interaction in the graph by setting the interaction as a root interaction, identifying a subtree interaction of a first depth from the root interaction and calculating a first discounted factor based on the first depth, identifying a subtree interaction of a second depth from the root interaction and calculating a second discounted factor based on the second depth, and wherein the engagement score is a sum of the first discounted factor multiplied by a first number of subtree interactions of the first depth and the second discounted factor multiplied by a second number of subtree interactions of the second depth;
weighting the interaction in the graph based on a type of interaction to drive users of the online service to engage with a particular type of interaction and for presentation of content items related to the particular type of interaction;
determining a network effect score for the first user interaction based on the engagement score and the weighting of the interaction in the graph;
generating and training, using the one or more processors, a model characterizing a network effect of the first user interaction with the content item in the online service based on the graph of interactions, different discount factors, and the network effect score;
generating, using the one or more processors, a predicted network effect score for the content item of the plurality of content items based on the model, the engagement score, and the weighting to predict the network effect made by presentation of the content item;
organizing, using the one or more processors, the plurality of content items based on the predicted network effect score; and
transmitting, using the one or more processors, the plurality of organized content items for presentation to a user.

2. The method of claim 1 , further comprising storing the content item and the engagement score in a data structure.

3. The method of claim 1 , further comprising:
associating, using the one or more processors, content item features with the network effect score for the first user interaction with the content item;
identifying, using the one or more processors, the content item features that are characteristic of a high network effect score for the first user interaction; and
wherein generating the model is based on identified content item features.

4. The method of claim 3 , wherein identifying the content item features that are characteristic of the high network effect score includes performing a regression analysis on the plurality of content items and associated network effect scores.

5. The method of claim 1 , wherein determining the network effect score for the first user interaction comprises:
determining, using the one or more processors, an interaction type to increase; and
increasing, using the one or more processors, a weight given to a selected user interaction in determining the network effect score.

6. The method of claim 1 , wherein the first user interaction is one from the group of: creating a new content item, endorsing an existing content item, resharing an existing content item, commenting on an existing content item, endorsing a comment on an existing content item, clicking-through a link in an existing content item, or playing media associated with an existing content item.

7. The method of claim 1 , comprising transmitting a recommended interaction for the content item to the user, wherein the recommended interaction is based on the predicted network effect score.

8. A computer program product comprising a non-transitory computer usable medium including a computer readable program, wherein the computer readable program when executed on a computer causes the computer to:
receive a record of data describing user engagement with a plurality of content items in an online service;
identify, from the record of data, a first user interaction with a content item and a sequence of interactions resulted from the first user interaction in the online service;
generate a graph of interactions with the content item in the online service, wherein a vertex of the graph represents a user interaction of identified interactions and an edge represents a causal relation between the identified interactions;
calculate an engagement score for an interaction in the graph by setting the interaction as a root interaction, identifying a subtree interaction of a first depth from the root interaction and calculating a first discounted factor based on the first depth, identifying a subtree interaction of a second depth from the root interaction and calculating a second discounted factor based on the second depth, and wherein the engagement score is a sum of the first discounted factor multiplied by a first number of subtree interactions of the first depth and the second discounted factor multiplied by a second number of subtree interactions of the second depth;
weight the interaction in the graph based on a type of interaction to drive users of the online service to engage with a particular type of interaction and for presentation of content items related to the particular type of interaction;
determine a network effect score for the first user interaction based on the engagement score and the weighting of the interaction in the graph;
generate and train a model characterizing a network effect of the first user interaction with the content item in the online service based on the graph of interactions, different discount factors, and the network effect score;
generate a predicted network effect score for the content item of the plurality of content items based on the model, the engagement score, and the weighting to predict the network effect made by presentation of the content item;
organize the plurality of content items based on the predicted network effect score; and
transmit the plurality of organized content items for presentation to a user.

9. The computer program product of claim 8 , wherein the computer readable program when executed on the computer causes the computer to store the content item and the engagement score in a data structure.

10. The computer program product of claim 8 , wherein the computer readable program when executed on the computer causes the computer to:
associate content item features with the network effect score for the first user interaction with the content item;
identify the content item features that are characteristic of a high network effect score for the first user interaction; and
wherein generating the model is based on identified content item features.

11. The computer program product of claim 10 , wherein to identify the content item features that are characteristic of the high network effect score, the computer readable program when executed on the computer causes the computer to perform a regression analysis on the plurality of content items and associated network effect scores.

12. The computer program product of claim 8 , wherein to determine the network effect score for the first user interaction, the comp

Assignees

• Google Inc
• Google Llc

Inventors

• Chandran Harish
• Teng Junbin
• Mazzia Alessandra

Classifications

• G06Q10/40

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

• G06F16/951 Primary

  Indexing; Web crawling techniques · CPC title

• G06F17/30864 Primary

  Physics · mapped topic

• G06Q30/0251 Primary

  Targeted advertisements · CPC title

• G06Q10/46

  Determination of level of influence of users within social networking services · CPC title