Predicting content performance with interest data

• US 9,549,017 B2
• Filed: 03/14/2013
• Issued: 01/17/2017
• Est. Priority Date: 08/29/2012
• Status: Active Grant

• Current Assignee: Google LLC (Alphabet Inc.)
• Original Assignee: Google Inc. (Alphabet Inc.)
• Inventors: Zhang, Xiaonan, Ponnekanti, Shankar, Zamir, Oren Eli, Liu, Ting
• Primary Examiner(s): Nano, Sargon
• Assistant Examiner(s): LI, MICHAEL J
• Application Number:
  US13/831,110
• Publication Number:
  US 20140068011A1
• Time in Patent Office: 1,405 Days
• US Class Current: 1/1
• CPC Class Codes: G06F 16/285 Clustering or classification
• CPC Class Codes: G06F 16/5838 using colour
• CPC Class Codes: G06F 16/5846 using extracted text
• CPC Class Codes: G06F 16/9535 Search customisation based on user profiles and personalisation
• CPC Class Codes: G06N 5/02 Knowledge representation; Symbolic representation
• CPC Class Codes: H04L 67/02 based on web technology, e.g. hypertext transfer protocol [HTTP]
• CPC Class Codes: H04L 67/303 Terminal profiles
• CPC Class Codes: H04L 67/53 using third party service providers

Abstract

Systems and methods for predicting content performance with interest data include receiving a content selection request that includes a client identifier. One or more topical interest categories associated with the client identifier may be used as inputs to a prediction model to predict the likelihood of an online action occurring as a result of third-party content being selected. The predicted likelihood may be used to select third-party content.

Claims

1. A method of selecting content, comprising:

receiving, at a processing circuit, a content selection request for third-party content to be presented by a client device in conjunction with first-party content, the content selection request including a client device identifier;

retrieving data indicative of a set of one or more topical interest categories associated with the client device identifier;

identifying, by the processing circuit, input values for a prediction model, the input values including;

the set of topical interest categories,

a click propensity metric associated with the client identifier, the click propensity metric indicative of how often the client device submitted clicks on previously selected third-party content, and

a length of time each topical interest category has been associated with the client device identifier, wherein a length of time greater than a defined time period indicates a long-term topical interest category and a length of time less than the defined time period indicates a short-term topical interest category;

performing, by the processing circuit, one or more of image recognition or text recognition on each of a set of third-party content to determine a topical interest category;

selecting, by the processing circuit, a subset of third-party content eligible for selection based in part on the set of topical interest categories associated with the client device identifier, a topical interest category of each of the subset of third-party content, and a topic of first party content;

using the prediction model that includes a logistic regression model to predict, based on the input values including the length of time, a predicted action metric for the third-party content in the subset, the predicted action metric corresponding to a predicted likelihood of an online action occurring as a result of the third-party content being selected, the logistic regression model indicating the predicted likelihood of the online action as

2. The method of claim 1, wherein the subset of third-party content eligible for selection is selected based in part on a match between a topic of the content in the subset and a topic of the first-party content.

3. The method of claim 1, wherein the predicted action metric is the predicted click through rate.

4. The method of claim 1, wherein the predicted action metric is the predicted conversion rate.

5. The method of claim 1, further comprising:

monitoring whether the predicted online action occurred; and

adjusting the prediction model based on the monitoring.

6. The method of claim 1, wherein the first-party content comprises an application executable by the client device.

7. The method of claim 1, wherein the indication comprises a command configured to cause the client device to retrieve the selected third-party content.

8. A system for selecting content, comprising a processing circuit operable to:

receive a content selection request for third-party content to be presented by a client device in conjunction with first-party content, the content selection request including a client device identifier;

retrieve data indicative of a set of one or more topical interest categories associated with the client device identifier;

identify input values for a prediction model, the input values including;

the set of topical interest categories,

a click propensity metric associated with the client identifier, the click propensity metric indicative of how often the client device submitted clicks on previously selected third-party content, and

a length of time each topical interest category has been associated with the client device identifier, wherein a length of time greater than a defined time period indicates a long-term topical interest category and a length of time less than the defined time period indicates a short-term topical interest category;

perform one or more of image recognition or text recognition on each of a set of third-party content to determine a topical interest category;

select a subset of third-party content eligible for selection based in part on the set of topical interest categories associated with the client device identifier, a topical interest category of each of the subset of third-party content, and a topic of first party content;

use the prediction model that includes a logistic regression model to predict, based on the input values including the length of time, a predicted action metric for the third-party content in the subset, the predicted action metric corresponding to a predicted likelihood of an online action occurring as a result of the third-party content being selected, the logistic regression model indicating the predicted likelihood of the online action as

9. The system of claim 8, wherein the subset of third-party content eligible for selection is selected based in part on a match between a topic of the content in the subset and a topic of the first-party content.

10. The system of claim 9, wherein the entity extractor comprises a closed captioning decoder, and wherein the audience measurement server is further configured to determine that the identified search query corresponds to the identified entity responsive to the search query being included in a closed captioning transcript of the item of broadcast content.

11. The system of claim 8, wherein the predicted action metric is the predicted conversion rate.

12. The system of claim 8, wherein the processing circuit is further configured to monitor whether the predicted online action occurred and adjust the prediction model based on the monitoring.

13. The system of claim 8, wherein the first-party content comprises an application executable by the client device.

14. A non-transitory computer-readable storage medium having machine instructions stored therein, wherein the instructions, when executed by a processor, cause the processor to perform operations comprising:

receiving a content selection request for third-party content to be presented by a client device in conjunction with first-party content, the content selection request including a client device identifier;

retrieving data indicative of a set of one or more topical interest categories associated with the client device identifier;

identifying input values for a prediction model, the input values including;

the set of topical interest categories,

a click propensity metric associated with the client identifier, the click propensity metric being indicative of how often the client device submitted clicks on previously selected third-party content, and

a length of time each topical interest category has been associated with the client device identifier, wherein a length of time greater than a defined time period indicates a long-term topical interest category and a length of time less than the defined time period indicates a short-term topical interest category;

performing, by the processing circuit, one or more of image recognition or text recognition on each of a set of third-party content to determine a topical interest category;

selecting a subset of third-party content eligible for selection based in part on the set of topical interest categories associated with the client device identifier, a topical interest category of each of the subset of third-party content, and a topic of first party content;

using the prediction model that includes a logistic regression model to predict, based on the input values including the length of time, a predicted action metric for the third-party content in the subset, the predicted action metric corresponding to a predicted likelihood of an online action occurring as a result of the third-party content being selected, the logistic regression model indicating the predicted likelihood of the online action as

15. The non-transitory computer-readable storage medium of claim 14, wherein the subset of third-party content eligible for selection is selected based in part on a match between a topic of the content in the subset and a topic of the first-party content.

16. The non-transitory computer-readable storage medium of claim 14, wherein the predicted action metric is the predicted click through rate.

17. The non-transitory computer-readable storage medium of claim 14, wherein the instructions stored therein further include instruction that, when executed by a processor, cause the processor to perform operations comprising:

monitoring whether the predicted online action occurred; and

adjusting the prediction model based on the monitoring.