US 12,278,796 B1
Context based user notification adjustment and management
Zachary A. Silverstein, Georgetown, TX (US); Jeremy R. Fox, Georgetown, TX (US); and Logan Bailey, Atlanta, GA (US)
Assigned to INTERNATIONAL BUSINESS MACHINES CORPORATION, Armonk, NY (US)
Filed by International Business Machines Corporation, Armonk, NY (US)
Filed on Oct. 24, 2023, as Appl. No. 18/492,926.
Int. Cl. G06F 15/16 (2006.01); G06F 40/35 (2020.01); H04L 51/043 (2022.01); H04L 51/226 (2022.01)
CPC H04L 51/226 (2022.05) [G06F 40/35 (2020.01); H04L 51/043 (2013.01)] 20 Claims
OG exemplary drawing
 
1. A computer-implemented method comprising:
intercepting, during a currently occurring interaction, a first notification intended for a user;
initially training a neural network model on notification data associated with a group of users;
further training, using machine learning, the neural network model with at least one previous interaction of the user, the further training producing a machine learning trained model;
causing, at least based on an analysis of data and metadata corresponding to the first notification, the machine learning trained model to produce an urgency level corresponding to the first notification;
causing, responsive to at least (i) a level of engagement of the user in the currently occurring interaction, (ii) an emotion/sentiment associated with the currently occurring interaction, and (iii) a comparison of the currently occurring interaction with the at least one previous interaction of the user, the machine learning trained model to produce an urgency level corresponding to the currently occurring interaction;
identifying, by analyzing the currently occurring interaction including the user, a lull in the currently occurring interaction, the lull comprising an amount of time in which the urgency level of the first notification becomes greater than the urgency level of the currently occurring interaction;
delivering, during the lull, the first notification to the user;
receiving information associated with a delivery of the first notification, the information including at least a user reaction to the first notification; and
adjusting the machine learning trained model based on the user reaction.