Back to results
US 2024/0403901 A1Published application
General Content Perception and Selection System
AssigneeImaige, Inc.
InventorsTroy DeBraal, Nethika Sahani Suraweera, Justin Williams
Life of this patent
Estimating the term…
Abstract
This invention is directed toward a system which can “step into the shoes” of a user and learn the perspective of that user, regarding photographs or other content, to the point where the system can learn, using criteria it has developed through its interaction with the user, to select photographs it predicts the user will find meaningful from large sets of photographs. The “meaningfulness” of various content from a multitude of users is a constantly improving system made up of four basic elements: a General Content Perspective, an Individual Content Perspective, a Natural Language Generation and Content Presentation, and a Hypersphere element.
Claims20
A computer-implemented method for video selection, executed by a hardware processor of a computer, said computer-implemented method comprising the steps of:
- having a user upload a group of videos from a computer device into a non-transitory computer-readable storage medium of the computer;
- having the computer store the group of videos in the non-transitory computer-readable storage medium;
- having the computer administer an explicit learning user survey to the user via a graphical user interface displayed on the computer device to obtain survey results;
- having the hardware processor of the computer perform an initial pixel analysis of the group of videos that have been uploaded and stored in the non-transitory computer-readable storage medium;
- having the hardware processor allocate an assigned persona to the user based on the initial pixel analysis of the group of videos that have been uploaded and the survey results;
- having the hardware processor use the assigned persona to serve as a baseline behavioral and preference model and having the hardware processor use the baseline behavioral and preference model to adjust predictions for content selection;
- having the hardware processor utilize a machine learning algorithm run by the hardware processor to create and train an initial individual model for the user, wherein the initial individual model comprises a user-specific predictive model tailored to understand subjective tendencies influenced by broad motivational biases, said initial individual model being developed through pixel-level analysis of the group of videos that have been uploaded, targeting identification of both human-readable and machine-readable behaviors, biases, and preferences encapsulated in the images during a video capture process;
- using said initial individual model in combination with a video evaluation persona and a universal video understanding model to select a set of videos from the group of videos that were uploaded, thereby providing a video selection;
- employing the graphical user interface on the computer device to present the video selection to the user;
- having the hardware processor generate natural language feedback via a natural language processing algorithm, said natural language feedback explaining a rationale behind the video selection to the user;
- employing the graphical user interface on the computer device to display the video selection and the natural language feedback;
- having the computer receive user feedback via the graphical user interface;
- having the hardware processor modify the initial individual model to create a modified individual model based on the received user feedback and further implicit learning from video analysis, using the machine learning algorithm;
- having the hardware processor re-select the set of videos for one of two purposes:
- (1) refining the video selection based on the modified individual model, or (2) selecting an alternative set from a new group of uploaded videos, both influenced by the modified individual model, wherein the modified individual model has an impact on both data generated by the machine learning algorithm and final selection criteria employed by the machine learning algorithm;
- displaying the re-selected set of videos to the user via the computer device;
- having the hardware processor monitor new content uploaded into the non-transitory computer-readable storage medium;
- and
- having the hardware processor reassess and adjust the modified individual model and the assigned persona based on the new content and further implicit learning from the video analysis, using the machine learning algorithm.
16 more elements
A computer-implemented method for video selection, executed by a hardware processor of a computer, said computer-implemented method comprising the steps of:
- having a user upload a plurality of content from a computer device into a non-transitory computer-readable storage medium of the computer;
- having the computer store the plurality of content in the non-transitory computer-readable storage medium;
- having the computer administer an explicit learning user survey to the user via a graphical user interface displayed on the computer device to obtain survey results;
- having the hardware processor of the computer perform an initial pixel analysis of the plurality of content that have been uploaded and stored in the non-transitory computer-readable storage medium;
- having the hardware processor allocate an assigned persona to the user based on the initial pixel analysis of the plurality of content that have been uploaded and the survey results;
- having the hardware processor use the assigned persona to serve as a baseline behavioral and preference model and having the hardware processor use the baseline behavioral and preference model to adjust predictions for content selection;
- having the hardware processor utilize a machine learning algorithm run by the hardware processor to create and train an initial individual model for the user, wherein the initial individual model comprises a user-specific predictive model tailored to understand subjective tendencies influenced by broad motivational biases, said initial individual model being developed through transforming pixel-level analysis of images that have been input targeting identification of patterns and objective and subjective learnings across the plurality of content that infer behaviors, biases, and preferences;
- using said initial individual model in combination with a video evaluation persona and a universal video understanding model to select a set of videos from the plurality of content that were uploaded, thereby providing a video selection;
- employing the graphical user interface on the computer device to present the video selection to the user;
- having the hardware processor generate natural language feedback via a natural language processing algorithm, said natural language feedback explaining a rationale behind the video selection to the user;
- employing the graphical user interface on the computer device to display the video selection and the natural language feedback;
- having the computer receive user feedback via the graphical user interface;
- having the hardware processor modify the initial individual model to create a modified individual model based on the received user feedback and further implicit learning from video analysis, using the machine learning algorithm;
- having the hardware processor re-select the set of videos for one of two purposes:
- (1) refining the video selection based on the modified individual model, or (2) selecting an alternative set from a new group of uploaded videos, both influenced by the modified individual model, wherein the modified individual model has an impact on both data generated by the machine learning algorithm and final selection criteria employed by the machine learning algorithm;
- displaying the re-selected set of videos to the user via the computer device;
- having the hardware processor monitor new content uploaded into the non-transitory computer-readable storage medium;
- and
- having the hardware processor reassess and adjust the modified individual model and the assigned persona based on the new content and further implicit learning from the video analysis, using the machine learning algorithm.
16 more elements
Dependencies are inferred from the claim text.