US 12,293,285 B2
Utilizing a neural network model to predict content memorability based on external and biometric factors
Edouard Mathon, Antibes (FR); Christian Souche, Cannes (FR); and Ji Tang, Valbonne (FR)
Assigned to Accenture Global Solutions Limited, Dublin (IE)
Filed by Accenture Global Solutions Limited, Dublin (IE)
Filed on May 6, 2021, as Appl. No. 17/302,569.
Prior Publication US 2022/0358357 A1, Nov. 10, 2022
Int. Cl. G06Q 30/02 (2023.01); G06F 3/01 (2006.01); G06F 11/34 (2006.01); G06N 3/044 (2023.01); G06N 3/08 (2023.01); G06N 7/01 (2023.01); G06Q 30/0251 (2023.01)
CPC G06N 3/08 (2013.01) [G06F 3/015 (2013.01); G06F 11/3438 (2013.01); G06N 3/044 (2023.01); G06N 7/01 (2023.01); G06Q 30/0255 (2013.01)] 19 Claims
OG exemplary drawing
 
1. A method to predict content memorability in a network, comprising:
receiving, by one or more processors of a device including hardwired circuitry, target user category data identifying a target user, daily user data associated with the target user, real-time user data associated with the target user, and content data, wherein the real-time user data includes biometric information of the target user, and wherein the biometric information of the target user is received from a sensor employed in one or more user devices;
converting, by the one or more processors, the target user category data and the daily user data into embedding formats to generate embedded target user category data and embedded daily user data;
normalizing, by the one or more processors, the real-time user data by using a time series matrix to generate normalized real-time user data;
converting, by the one or more processors, the normalized real-time user data into embedded real-time user data by processing the normalized real-time user data with one or more neural network models, wherein the one or more neural network models includes at least one of a recurrent neural network model, a long short-term memory neural network model, or a transformer neural network model;
converting, by the one or more processors, the content data into embedded content data by:
converting one or more images provided in the content data into the embedded content by processing the one or more images with a residual neural network model;
converting one or more audio files provided in the content data into the embedded content by processing the one or more audio files with a Mobilenet Network model; and
converting textual information provided in the content data into the embedded content by processing the textual information with a sentence transformer model;
processing, by the one or more processors, the embedded target user category data, the embedded daily user data, and the embedded real-time user data, with the one or more neural network models, to determine a real-time user state;
determining, by the one or more processors, a real-time user memory score based on the real-time user state;
processing, by the one or more processors, the embedded content data, the real-time user state, and the real-time user memory score, with the one or more neural network models, to determine a memorability score for the content data and the target user to predict the content memorability in the network; and
performing, by the one or more processors, one or more actions based on the memorability score, wherein the one or more actions comprise retraining the one or more neural network models based on the memorability score.