Walmart Apollo, LLC

Wal-mart Stores, Inc.

Omkar Deshpande

Kannan Achan

Abhimanyu Mitra

Afroza Ali

Abhay Kashyap

Yang Deng

Digvijay Singh Lamba

Namrata Pramodkumar Tholiya

Systems and methods for content selection and presentation are disclosed. Training data including indicating one or more interactions with one or more content elements and associated with one of a plurality of individual contexts is received. A selection model is trained by applying a reinforcement learning mechanism and an individual explore-exploit mechanism. A context for a user is selected by applying the selection model, which is configured to determine an expected future reward value of at least one of the plurality of individual contexts, determine an expected future reward value of a global context based on a past click-through rate, a reward value, and a future click-through rate, and, select the global context or one of the one or more individual contexts based on a comparison of the expected future reward values.

Dynamic evaluation and use of global and contextual personas

A system and method for content selection and presentation is disclosed. A plurality of content elements configured for presentation in at least one content container is received and one of the plurality of content elements is selected for presentation in the at least one content container. The one of the plurality of content elements is selected by a trained selection model configured to select one of an individual context or a global context. An interface including the selected one of the plurality of content elements is generated.

A system including one or more processors and one or more non-transitory computer-readable media storing computing instructions configured to run on the one or more processors and perform: displaying content elements on one or more websites to users; tracking respective impression response data for each impression of a respective content element of the content elements comprising (a) a respective response of a respective user of the users and (b) a respective time of the respective response of the respective user; determining respective weightings of the content elements based on posterior distributions using the respective impression response data, as adjusted by a temporal decay factor, based on the respective times of the respective impression response data for the content elements; and generating a webpage of the one or more web sites to comprise a selected content element based on the respective weighting of the selected content element. Other embodiments are disclosed.

Automated mechanisms to resolve explore-exploit dilemma with adaptive revival opportunities

A system including one or more processors and one or more non-transitory computer-readable media storing computing instructions configured to run on the one or more processors and perform: displaying content elements on one or more websites to users; performing a classification of the users into segments; receiving a request from a first user of the users to display a webpage of the one or more websites; determining an amount of information in a respective posterior distribution of each of the one or more first segments of the segments in which the first user is classified, for each of the content elements based on impression response data; selecting a selected content element from among the content elements based on weightings of the content elements for the one or more first segments; and generating the webpage comprising the selected content element. Other embodiments are disclosed.

Personalized mechanisms to resolve explore-exploit dilemma with dynamically shared learnings

A method including displaying content elements on one or more websites to users. This method can also include tracking impression response data comprising (a) a response of the user to the content element of the content elements displayed on the one or more websites and (b) a time of the response of the user. Additionally, the method can include receiving a request from a first user. The method can further include determining weightings of the content elements based on posterior distributions using the impression response data, as adjusted by a temporal decay factor based on the times of the impression response data for the content elements. The method additionally can include generating the webpage to comprise the selected content element. Other embodiments are disclosed.

A method including displaying content elements on one or more websites to users. The classification of the users into segments can be based on each impression of content elements being displayed on the one or more websites to a user of the users, tracking impression response data comprising (a) a response of the user to the content element of the content elements displayed on the one or more websites, and (b) one or more segments of the segments in which the user is classified. The method can also include receiving a request from a first user. The method can also include generating a mixture distribution for the first segment for the first content element based on the impression response data for the first segment and generating the webpage to comprise the selected content element. Other embodiments are disclosed.

A method for merging two taxonomies is disclosed. Top levels of first and second taxonomies are merged. For the second taxonomy nodes are evaluated and selectively merged with nodes of the first taxonomy according to comparison of scores for these nodes with a threshold. The score for a node of the first taxonomy is a combination of one or more of a lineage quality score, Jaccard distance, string edit distance, and category depth score. After an iteration, mergings between nodes of the first and second taxonomies may be reversed if child nodes of the merged nodes were not likewise merged. Iterations may be repeated until no nodes are merged in an iteration.

Automatic taxonomy merge

Methods are disclosed for converting a directed graph to a taxonomy using guidelines from a user. An initial tree is output from a first pruning step in which subtree preferences (and other weights) are applied to preserve or remove paths from a node to one or more levels of descendent nodes. Subtree preferences (and infoboxes) may specify rules for automatically generating recommendations during application to nodes. In a second pruning step, the directed graph is again processed with additional weightings applied to edges in the graph in accordance with the recommendations. The recommendations may be human defined. Recommendations may specify a recommended ancestor for a particular node and may include a weighting to be applied to the recommendation itself, if there are multiple conflicting recommendations for the same node. Recommendations may also specify what standard weight to apply to the edge of the best parent.

Growing community of inventors

Saratoga, CA, United States of America

Omkar Deshpande