US 12,008,826 B2
Method and apparatus for customized deep learning-based text correction
Junchao Wei, San Mateo, CA (US)
Assigned to KONICA MINOLTA BUSINESS SOLUTIONS U.S.A., INC., Foster City, CA (US)
Filed by KONICA MINOLTA BUSINESS SOLUTIONS U.S.A., INC., Foster City, CA (US)
Filed on Sep. 30, 2021, as Appl. No. 17/491,122.
Prior Publication US 2023/0096700 A1, Mar. 30, 2023
Int. Cl. G06V 30/12 (2022.01); G06F 18/20 (2023.01); G06F 18/214 (2023.01); G06N 3/045 (2023.01); G06V 30/18 (2022.01)
CPC G06V 30/12 (2022.01) [G06F 18/214 (2023.01); G06F 18/285 (2023.01); G06N 3/045 (2023.01); G06V 30/18 (2022.01)] 20 Claims
OG exemplary drawing
 
1. A computer-implemented text correction method comprising:
with a processor, providing one of base data and end user data to pretrain a base deep learning model or an end user deep learning model, respectively, as a pretrained base deep learning model or a pretrained end user deep learning model;
providing a processor-based data mode classifier to select one of the pretrained base deep learning model or a pretrained end user deep learning model;
responsive to an input dataset, selecting one of the pretrained base deep learning model or the processor-based data mode classifier;
responsive to an output of the data mode classifier, performing training on one of the base deep learning model or the end user deep learning model;
responsive to a size of the dataset, performing retraining on one of the base deep learning model or the end user deep learning model;
responsive to the training of one of the base deep learning model or the end user deep learning model, providing outputs signifying a confidence level of the training; and
responsive to the confidence level of the training, providing output text to an end user.