US 12,229,152 B2
Systems and methods of dynamic page layout using exploration-exploitation models
Afroza Ali, Los Altos, CA (US); Zhihao Huang, San Jose, CA (US); Abhimanyu Mitra, Cupertino, CA (US); Atul Kochhar, Karnataka (IN); and Kannan Achan, Saratoga, CA (US)
Assigned to Walmart Apollo, LLC, Bentonville, AR (US)
Filed by Walmart Apollo, LLC, Bentonville, AR (US)
Filed on Jan. 31, 2023, as Appl. No. 18/104,112.
Prior Publication US 2024/0256556 A1, Aug. 1, 2024
Int. Cl. G06F 16/2457 (2019.01); G06F 9/451 (2018.01); G06F 16/215 (2019.01); G06F 16/248 (2019.01)
CPC G06F 16/24578 (2019.01) [G06F 9/451 (2018.02); G06F 16/215 (2019.01); G06F 16/24575 (2019.01); G06F 16/248 (2019.01)] 20 Claims
OG exemplary drawing
 
1. A system comprising:
a processor; and
a non-transitory memory storing instructions that, when executed, cause the processor to:
generate a first interface including a first set of content modules comprising a set of generic content modules and a set of dynamic content modules selected from a set of candidate content modules based on a set of ranking weights, wherein each content module of the set of content modules includes one or more items including a link to navigate or obtain a function from the first interface;
receive feedback data representative of one or more interactions with the first interface, wherein the feedback data includes positive feedback and negative feedback;
generate a trained ranking model by inputting a training dataset including a set of contextual data, positive data, and negative data, wherein the training dataset is normalized by a normalization module, and wherein the trained ranking model is generated by iteratively revising a set of model parameters;
generate the set of ranking weights for each content module in the set of candidate content modules, wherein the ranking weights are generated by a contextual explore-exploit model based on the feedback data;
modify the inclusion of each of the dynamic content modules by adjusting the ranking weights for each content module in the set of content modules based on a variance control parameters, wherein the modification is performed by an exploration module;
receive a request for a second interface including at least a user identifier and a plurality of contexts, wherein each of the plurality of contexts are based on at least one of an event category, a general interest category, or a personalized interest category;
generate a first user segmentation based on at least one of the user identifier, a first context of the plurality of contexts, and historical user data;
generate a second segmentation based on at least one of the user identifier, a second context of the plurality of contexts, and historical user data;
generate, by the trained ranking model, a set of ranked content modules for the first context and the second context, wherein the trained ranking model receives a corresponding user segmentation and a rank of dynamic content, wherein the rank of dynamic content modules is generated based on the user segmentation and the set of ranking weights;
generate a second set of content modules for the first segmentation and a third set of content modules for the second segmentation;
generate a second interface including the second set of content modules and the third set of content modules, wherein the second set of content modules and the third set of content modules includes a predetermined number of top ranked content modules in the set of ranked content modules; and
provide the generated second interface to a user system, wherein a user or operator of the user system may interact with the generated second interface.