US 12,033,184 B2
Digital channel personalization based on artificial intelligence (AI) and machine learning (ML)
Marc Perreau Guimaraes, Los Gatos, CA (US); Tetiana Kostenko, Frederiksberg C (DK); Samira Sadeghi, San Francisco, CA (US); Mingde Xu, San Jose, CA (US); Abhishek Soni, Hayward, CA (US); Romeo B. Valencia, San Francisco, CA (US); and Nancy Huei-Jiun Lee, San Jose, CA (US)
Assigned to SITECORE CORPORATION A/S, Copenhagen (DK)
Filed by Sitecore Corporation A/S, Copenhagen (DK)
Filed on Oct. 30, 2020, as Appl. No. 17/085,680.
Prior Publication US 2022/0138798 A1, May 5, 2022
Int. Cl. G06Q 30/02 (2023.01); G06F 16/90 (2019.01); G06F 16/906 (2019.01); G06N 20/00 (2019.01); G06Q 30/0251 (2023.01)
CPC G06Q 30/0254 (2013.01) [G06F 16/906 (2019.01); G06N 20/00 (2019.01)] 24 Claims
OG exemplary drawing
 
1. A computer-implemented method for personalizing a digital channel, comprising:
(a) providing a digital channel to multiple users, wherein the digital channel comprises a website of a business;
(b) collecting visitor information at each visit of each of the multiple users to the website, wherein:
(1) the visitor information comprises data about each visit;
(2) the visitor information comprises session data, user-item data, and outcome data, wherein:
(i) the session data comprises contextual data about each visit;
(ii) the user-item data comprises events associating each of the multiple users at each visit to the contextual data;
(iii) the outcome data records events related to a business value of a session;
(iv) the outcome data comprises binary data or numerical monetary value data; and
(v) the events comprise actions by each of the multiple users that trigger a gain to the business;
(3) each visit comprises multiple content items that are presented via the website, and wherein each of the multiple content items comprises text, an image, or a video of the business; and
(4) prior to each visit, no past knowledge of the multiple users is known;
(c) autonomously clustering the multiple users, wherein:
(1) the clustering segments a user population into two or more behavioral groups based on the visitor information;
(2) the clustering maximizes mutual information between the multiple users in an assigned behavioral group and one or more of the multiple content items; and
(3) the two or more behavioral groups are not set up manually;
(d) based on the clustering, generating and training a model for an interaction between each of the multiple users and each of the multiple content items, wherein:
(1) the model estimates a score for each interaction;
(2) the model is updated at a defined interval based on the visits of the multiple users to the website;
(3) the model comprises a machine learning model;
(e) determining, using the model and based on the score, which of the multiple content items to recommend to a specific user of the multiple users, wherein the determining jointly maximizes an outcome and a learning speed of the model; and
(f) personalizing and delivering the website for and to the specific user based on the recommended multiple content items, wherein the personalizing comprises components of the website rendering a variety of the multiple content items optimized for the specific user.