Back to results
US 11,153,655 B1Granted
Content appeal prediction using machine learning
AssigneeAmazon Technologies, Inc.
InventorsVernon Germano, Kripa Kanchana Sivakumar, Xiang Hao
Life of this patent
Estimating the term…
Abstract
Methods and apparatus are relating to the use of machine learning techniques to identify unrated content that will be appealing to particular demographics. Using explicit feedback (e.g., star ratings) and implicit feedback (e.g., viewing behavior) for a given demographic, feature sets are extracted from the rated titles and then used to recognize similar titles in unrated content.
Claims20
A computer-implemented method, comprising:
- identifying a first plurality of content items, each of the first plurality of content items having a content score for a first demographic profile that exceeds an appeal threshold, each content score being based on interactions of a first plurality of users of the first demographic profile with the corresponding content item;
- generating a feature set for each of the first plurality of content items, including processing each of the first plurality of content items using a computer vision technique to identify first visual elements from video frames of the first plurality of content items, features in the feature sets for the first plurality of content items representing the first visual elements;
- generating a feature set for each of a second plurality of content items, including processing each of the second plurality of content items using the computer vision technique to identify second visual elements from video frames of the second plurality of content items, features in the feature sets for the second plurality of content items representing the second visual elements;
- applying weights to at least some of the features in each of the feature sets based on one or both of the first demographic profile and a content type of the corresponding content item;
- using a machine learning model, identifying a first content item from the second plurality of content items by comparing the feature sets for the first plurality of content items with the feature sets for the second plurality of content items;
- and
- designating the first content item for recommendation in connection with the first demographic profile.
4 more elements
A computer program product, comprising one or more non-transitory, computer-readable media having computer program instructions stored therein, the computer program instructions being configured such that, when executed by one or more computing devices, the computer program instructions cause the one or more computing devices to:
- identify a first plurality of content items, each of the first plurality of content items having a content score for a first demographic profile that exceeds an appeal threshold, each content score being based on interactions of a first plurality of users of the first demographic profile with the corresponding content item;
- generate a feature set for each of the first plurality of content items, including processing each of the first plurality of content items using a computer vision technique to identify first visual elements from video frames of the first plurality of content items, features in the feature sets for the first plurality of content items representing the first visual elements;
- generate a feature set for each of a second plurality of content items, including processing each of the second plurality of content items using the computer vision technique to identify second visual elements from video frames of the second plurality of content items, features in the feature sets for the second plurality of content items representing the second visual elements;
- apply weights to at least some of the features in each of the feature sets based on one or both of the first demographic profile and a content type of the corresponding content item;
- use a machine learning model to identify a first content item from the second plurality of content items by comparing the feature sets for the first plurality of content items with the feature sets for the second plurality of content items;
- and
- designate the first content item for recommendation in connection with the first demographic profile.
4 more elements
A system, comprising one or more computing devices including one or more processors and memory configured to:
- identify a first plurality of content items, each of the first plurality of content items having a content score for a first demographic profile that exceeds an appeal threshold, each content score being based on interactions of a first plurality of users of the first demographic profile with the corresponding content item;
- generate a feature set for each of the first plurality of content items, including processing each of the first plurality of content items using a computer vision technique to identify first visual elements from video frames of the first plurality of content items, features in the feature sets for the first plurality of content items representing the first visual elements;
- generate a feature set for each of a second plurality of content items, including processing each of the second plurality of content items using the computer vision technique to identify second visual elements from video frames of the second plurality of content items, features in the feature sets for the second plurality of content items representing the second visual elements;
- apply weights to at least some of the features in each of the feature sets based on one or both of the first demographic profile and a content type of the corresponding content item;
- use a machine learning model to identify a first content item from the second plurality of content items by comparing the feature sets for the first plurality of content items with the feature sets for the second plurality of content items;
- and
- designate the first content item for recommendation in connection with the first demographic profile.
4 more elements
Dependencies are inferred from the claim text.