Predicting digital personas for digital-content recommendations using a machine-learning-based persona classifier

Patent metadata

Field	Value
Publication number	US-10904599-B2
Application number	US-201815994776-A
Country	US
Kind code	B2
Filing date	May 31, 2018
Priority date	May 31, 2018
Publication date	Jan 26, 2021
Grant date	Jan 26, 2021

Abstract

This disclosure relates to methods, non-transitory computer readable media, and systems that determine multiple personas corresponding to a user account for digital content and train a persona classifier to predict a given persona (from among the multiple personas) for content requests associated with the user account. By using the persona classifier, the disclosed methods, non-transitory computer readable media, and systems accurately detect a given persona for a content request upon initiation of the request. Based on determining the given persona, in some implementations, the methods, non-transitory computer readable media, and systems generate a digital-content recommendation for presentation on a client device associated with the user account.

First claim

We claim:

1. In a digital medium environment for detecting digital personas corresponding to user accounts, a computer-implemented method of training and applying persona classifiers comprising:
generating a plurality of initial bins corresponding to a plurality of projection values representing a plurality of content-consumption events of a user account;
creating a plurality of persona bins from the plurality of initial bins by combining initial bins together or maintaining initial bins apart based on projection-value differences among the plurality of initial bins;
detecting a content request from a client device of a plurality of client devices corresponding to the user account;
determining contextual features for the content request by identifying a location identifier for a location of the client device and a computing-device identifier for the client device from among a plurality of computing-device identifiers corresponding to the user account;
utilizing a persona classifier to determine a persona for the content request from among a plurality of personas associated with the user account based on the contextual features by mapping the contextual features to a persona bin from among the plurality of persona bins; and
utilizing a factorization machine to generate, for display on the client device, a digital-content recommendation based on the persona.

2. The method of claim 1 , wherein the plurality of personas associated with the user account comprise:
a first persona corresponding to a first combination of one or more of a computing device, a genre of digital content, a location, or a time of day; and
a second persona corresponding to a second combination of one or more of a computing device, a genre of digital content, a location, or a time of day.

3. The method of claim 1 , wherein the plurality of content-consumption events comprises one or more of user ratings of digital content associated with the user account or content-consumption sessions of digital content that progress to a threshold time within the digital content.

4. The method of claim 1 , wherein utilizing the factorization machine to generate the digital-content recommendation comprises generating a recommendation for digital-audio content or digital-video content.

5. The method of claim 1 , further comprising providing the digital-content recommendation to the client device by:
providing the digital-content recommendation as a selectable option to transmit digital content to the client device; or
providing the digital content for the client device to present within a graphical user interface and without user selection of the digital content.

6. A non-transitory computer readable storage medium comprising instructions that, when executed by at least one processor, cause a system to:
generate a plurality of initial bins corresponding to a plurality of projection values representing a plurality of content-consumption events of a user account;
create a plurality of persona bins from the plurality of initial bins by combining initial bins together or maintaining initial bins apart based on projection-value differences among the plurality of initial bins;
receive a content request from a client device of a plurality of client devices corresponding to the user account;
determine contextual features for the content request by identifying a location identifier for a location of the client device and a computing-device identifier for the client device from among a plurality of computing-device identifiers corresponding to the user account;
utilize a persona classifier to determine a persona for the content request from among a plurality of personas corresponding to the user account based on the contextual features by mapping the contextual features to a persona bin from among the plurality of persona bins; and
generate, for display on the client device, a digital-content recommendation based on the persona.

7. The non-transitory computer readable storage medium of claim 6 , further comprising instructions that, when executed by the at least one processor, cause the system to determine the contextual features for the content request after the client device transmits the content request and before transmission of digital content to the client device.

8. The non-transitory computer readable storage medium of claim 6 , wherein the plurality of personas corresponding to the user account comprise:
a first persona corresponding to a first preference for one or more first genres of digital content and a first combination of one or more computing devices, one or more locations, and one or more times; and
a second persona corresponding to a second preference for one or more second genres of digital content and a second combination of one or more computing devices, one or more locations, and one or more times.

9. The non-transitory computer readable storage medium of claim 6 , further comprising instructions that, when executed by the at least one processor, cause the system to determine the contextual features for the content request by identifying a time indicator for a time of day.

10. The non-transitory computer readable storage medium of claim 6 , further comprising instructions that, when executed by the at least one processor, cause the system to generate the digital-content recommendation based on the persona by utilizing a factorization machine or a regression model to generate the digital-content recommendation based on the persona.

11. A system for applying persona classifiers to predict personas for content requests comprising:
at least one processor;
at least one non-transitory computer memory comprising a persona classifier, a factorization machine, a content-consumption log for a user account, and instructions that, when executed by at least one processor, cause the system to:
analyze the content-consumption log comprising a plurality of content-consumption events for the user account;
generate a plurality of projection values representing the plurality of content-consumption events of the user account;
train the persona classifier to predict personas for content requests by:
generating a plurality of initial bins corresponding to the plurality of projection values;
creating a plurality of persona bins from the plurality of initial bins by combining initial bins together or maintaining initial bins apart based on projection-value differences among the plurality of initial bins; and
generating a persona-prediction tree that maps training contextual features to the plurality of persona bins, wherein the training contextual features correspond to the plurality of content-consumption events and comprise location identifiers for locations of client devices corresponding to the user account and computing-device identifiers for client devices corresponding to the user account.

12. The system of claim 11 , wherein each content-consumption event is represented by a feature vector and each feature vector representing a content-consumption event comprises contextual-feature values corresponding to one or more of a computing device, a genre of digital content, a location, or a time of day.

13. The system of claim 11 , further comprising instructions that, when executed by the at least one processor, cause the system to:
generate a principal-component-analysis projection based on a plurality of feature vectors representing the plurality of content-consumption events; and
generate the plurality of projection values representing the plurality of content-consumption events based on the principal-component-analysis projection.

14. The system of claim 11 ,

Assignees

• Adobe Inc

Inventors

• Sarkhel Somdeb
• Swaminathan Viswanathan
• Yang Shuo
• Mitra Saayan
• Shivalingaiah Lakshmi
• Boyer Jason
• Rodgers Dwight

Classifications

• G06N20/00 Primary

  Machine learning · CPC title

• H04N21/251 Primary

  Learning process for intelligent management, e.g. learning user preferences for recommending movies (details of learning user preferences for the retrieval of video data in a video database G06F16/739; computer systems using learning methods G06N3/08) · CPC title

• G06F18/23213

  with fixed number of clusters, e.g. K-means clustering · CPC title

• G06F18/231

  Hierarchical techniques, i.e. dividing or merging pattern sets so as to obtain a dendrogram · CPC title

• G06F18/2135

  based on approximation criteria, e.g. principal component analysis · CPC title