Using polling results as discrete metrics for content quality prediction model

• US 9,582,812 B2
• Filed: 04/15/2014
• Issued: 02/28/2017
• Est. Priority Date: 04/07/2011
• Status: Active Grant

• Current Assignee: Meta Platforms, Inc. (f/k/a Facebook, Inc.)
• Original Assignee: Meta Platforms, Inc. (f/k/a Facebook, Inc.)
• Inventors: Yan, Rong, Hegeman, John
• Primary Examiner(s): SALAD, ABDULLAHI ELMI
• Application Number:
  US14/253,138
• Publication Number:
  US 20140229234A1
• Time in Patent Office: 1,050 Days
• Field of Search: 709/204, 709217-219, 707/748, 705/14.41, 705/14.42, 705/14.44, 705/14.43
• US Class Current: 1/1
• CPC Class Codes: G06N 5/048 Fuzzy inferencing
• CPC Class Codes: G06Q 30/02 Marketing; Price estimation or determination; Fundraising
• CPC Class Codes: G06Q 30/0202 Market predictions or forecasting for commercial activities
• CPC Class Codes: G06Q 30/0245 Surveys

Abstract

A social networking system presents content items to users, who then provide feedback regarding pairs of content items. The feedback includes a selection of a content item of the pair of content items that was preferred by the user over the other content item. The social networking system uses this information to train a predictive model that scores content items based on quality. The content items may be advertisements. The social networking system uses the pair-wise comparisons of the advertisements to determine feedback coefficients in an advertising quality score prediction model using regression analysis of the pair-wise comparisons for each predictive factor in the model. In this way, the pair-wise comparisons are used to train the prediction model to understand which advertisements are more enjoyable than others. A feedback coefficient for each predictive factor may be computed based on the preferences received from the group of users.

Claims

1. A computer-implemented method comprising:

determining interests for each of a plurality of users of a social networking system, a respective interest for each user based on information provided by the respective user;

inferring a preference of a user of the plurality of users for a first content item over a second content item, the inferring predicting, based on at least one of the determined interests for the user, which of the first and second content items the user would choose in a pair-wise comparison of the first and second content items;

identifying at least one predictive factor associated with the inferred preference for the first content item over the second content item;

for each identified predictive factor, determining a feedback coefficient for the identified predictive factor based on the inferred preference; and

storing the feedback coefficients for the identified predictive factors in a computer-readable storage medium.

2. The computer-implemented method of claim 1, wherein determining the feedback coefficient comprises:

analyzing the inferred preference using a statistical model; and

determining the feedback coefficient as a result of the analyzing.

3. The computer-implemented method of claim 2, wherein determining the feedback coefficient further comprises:

analyzing the inferred preference using a logistic regression; and

determining the feedback coefficient as a result of the analyzing.

4. The computer-implemented method of claim 1, further comprising:

receiving new content items for display to a user;

generating a quality score for each of the new content items based on a combination of the stored feedback coefficients for the identified predictive factors and profile information about the user; and

ranking the new content items for display based on the generated quality scores.

5. The computer-implemented method of claim 1, wherein the content items include advertisements for display to users of the social networking system.

6. The computer-implemented method of claim 1, wherein the content items include photos shared by users of the social networking system.

7. The computer-implemented method of claim 1, wherein the content items include music shared by users of the social networking system.

8. The computer-implemented method of claim 1, wherein the content items include videos shared by users of the social networking system.

9. The computer-implemented method of claim 1, wherein the content items include specified users, links to external websites, fan pages, groups of users, or interests shared by users of the social networking system.

10. The computer-implemented method of claim 1, wherein inferring a preference comprises:

receiving quantifiable scores of a plurality of content items; and

for each of the plurality of content items, generating a unique pair-wise comparison of two of the plurality of content items by ranking the content items by quantifiable scores associated with the content items, the ranking indicating the inferred preference of the user for the first content item over the second content item.

11. The computer-implemented method of claim 1, wherein the first and second content items are advertisements, the method further comprising:

for each advertisement available for display to a user, determining a quality perception score for the user by combining the feedback coefficients and determined interests for the user;

providing for display to the user a quality advertisement based on the quality perception score associated with the quality advertisement.

12. A non-transitory computer-readable storage medium storing instructions, the instructions when executed by a processor in a social networking system for predicting quality of content items, causes the processor to:

determine interests for each of a plurality of users of a social networking system, a respective interest for each user based on information provided by the respective user;

infer a preference of a user of the plurality of users for a first content item over a second content item, the inferring predicting, based on at least one of the determined interests for the user, which of the first and second content items the user would choose in a pair-wise comparison of the first and second content items;

identify at least one predictive factor associated with the inferred preference for the first content item over the second content item;

for each identified predictive factor, determine a feedback coefficient for the identified predictive factor based on the inferred preference; and

store the feedback coefficients for the identified predictive factors in a computer-readable storage medium.

13. The computer-readable storage medium of claim 12, wherein determining the feedback coefficient comprises:

analyzing the inferred preference using a statistical model; and

determining the feedback coefficient as a result of the analyzing.

14. The computer-readable storage medium of claim 13, wherein determining the feedback coefficient further comprises:

analyzing the inferred preference using a logistic regression; and

determining the feedback coefficient as a result of the analyzing.

15. The computer-readable storage medium of claim 12, wherein the executed instructions further cause the processor to:

receive new content items for display to a user;

generate a quality score for each of the new content items based on a combination of the stored feedback coefficients for the identified predictive factors and determined interests for the user; and

rank the new content items for display based on the generated quality scores.

16. The computer-readable storage medium of claim 12, wherein the content items include advertisements for display to users of the social networking system.

17. The computer-readable storage medium of claim 12, wherein the content items include links to external websites shared by users of the social networking system.

18. A system comprising:

a processor;

a non-transitory computer readable storage medium for storing software modules; and

a polling analysis module configured to;

determine interests for each of a plurality of users of a social networking system, a respective interest for each user based on information provided by the respective user;

infer a preference of a user of the plurality of users for a first content item over a second content item, the inferring predicting, based on at least one of the determined interests for the user, which of the first and second content items the user would choose in a pair-wise comparison of the first and second content items;

identify at least one predictive factor associated with the inferred preference for the first content item over the second content item;

for each identified predictive factor, determine a feedback coefficient for the identified predictive factor based on the inferred preference; and

store the feedback coefficients for the identified predictive factors in a computer-readable storage medium.

19. The system of claim 18, wherein determining the feedback coefficient comprises:

analyzing the inferred preference using a statistical model; and

determining the feedback coefficient as a result of the analyzing.

20. The system of claim 19, wherein determining the feedback coefficient further comprises:

analyzing the inferred preference using a logistic regression; and

determining the feedback coefficient as a result of the analyzing.

21. The system of claim 18, wherein the polling analysis module is further configured to:

receive new content items for display to a user;

generate a quality score for each of the new content items based on a combination of the stored feedback coefficients for the identified predictive factors and determined interests for the user; and

rank the new content items for display based on the generated quality scores.

22. The system of claim 18, wherein the content items include advertisements for display to users of the social networking system.

23. The system of claim 18, wherein the content items include links to external websites shared by users of the social networking system.