Method and apparatus for automatically providing advertisements

Patent metadata

Field	Value
Publication number	US-11928709-B2
Application number	US-201816195797-A
Country	US
Kind code	B2
Filing date	Nov 19, 2018
Priority date	Nov 19, 2018
Publication date	Mar 12, 2024
Grant date	Mar 12, 2024

Abstract

This application relates to apparatus and methods for determining data outputs to advertise on a platform such as a website. A computing device receives a website search request and determines a search term keyword. The computing device also determines a plurality of item accounts, such as sponsor campaigns, based on the search term keyword and a corresponding keyword of each item account. Each item account also includes a corresponding data output, such as a digital advertisement. The computing device identifies one of the item accounts based on determining an engagement probability for the digital advertisement of each item account. The engagement probability is determined based on aggregated impression and engagement data for the digital advertisement. The computing device provides the digital advertisement for the identified item account to a server, which may display the digital advertisement on the platform.

First claim

What is claimed is:

1. A system comprising:
a server configured to:
generate, based on an exponentially time decayed algorithm, aggregated impression data by monitoring a frequency of display of each of a plurality of elements in one or more prior interfaces over a specified period of time, wherein the aggregated impression data is stored in a database;
generate, based on the exponentially time decayed algorithm, aggregated engagement data by monitoring a frequency of interaction for each of the plurality of elements in the one or more prior interfaces over a specified period of time, wherein the aggregated engagement data is stored in the database;
receive a request for an interface from a requesting system, the request including a search request;
generate the interface including a corresponding item associated with a selected campaign; and
provide the interface including the corresponding item to the requesting system that generated the request for the interface, wherein the requesting system is configured to display the interface;

a keyword determination engine configured to:
receive, from the server, the search request;
extract search terms from the search request by applying a natural language processing algorithm to the search request;
identify at least one search keyword by parsing the search terms;
determine a plurality of campaigns, wherein each campaign of the plurality of campaigns includes data elements representative of at least one item keyword and at least one corresponding item, wherein the plurality of campaigns includes campaigns having at least one item keyword corresponding to the at least one search keyword; and

a processor configured to:
determine an overall engagement probability for each of the plurality of campaigns, wherein the overall engagement probability is determined by:
generating an item engagement probability distribution by a posterior probability distribution of the aggregated impression data and the aggregated engagement data of the at least one corresponding item, wherein the posterior probability distribution is based on a binomial and a beta distribution;
generating, by a machine learning model and based on the item engagement probability distribution, an item engagement probability for the at least one corresponding item of each of the plurality of campaigns; and
generating the overall engagement probability for each campaign by averaging the item engagement probability for each item of the campaign or selecting a maximum item engagement probability for the at least one corresponding item;

select the campaign of the plurality of campaigns having a highest overall engagement probability; and
provide output data to the server including the at least one corresponding item associated with the campaign;

wherein after the server generates the interface, the server is configured to:
receive impression data and engagement data for the interface including the at least one corresponding item associated with the selected campaign;
generate updated aggregated impression data and updated aggregated engagement data based on the received impression data and engagement data for the interface including the at least one corresponding item associated with the selected campaign; and
generate a second interface including at least one second item associated with a second campaign, wherein the second campaign is selected based on an updated overall engagement score of each campaign, wherein the updated overall engagement score is based, at least in part, on updated item engagement probabilities based on the updated aggregated impression data and the updated aggregated engagement data.

2. The system of claim 1 , wherein the updated aggregated impression data and the updated aggregated engagement data is generated by applying the exponentially time decayed algorithm to the received impression data and engagement data for the interface including the corresponding item associated with the selected campaign.

3. The system of claim 1 , wherein determining the overall engagement probability for each of the plurality of campaigns includes modifying the averaged item engagement probability or the maximum item engagement probability is based on a price per engagement for the at least one corresponding item for each of the plurality of campaigns.

4. The system of claim 1 , wherein determining the overall engagement probability includes:
generating an engagement probability for brand data for each of the plurality of campaigns; and
determining the overall engagement probability based on a summation of the engagement probability for the brand data and the item engagement probability of the at least one corresponding item for each of the campaigns.

5. The system of claim 1 , wherein the plurality of campaigns is determined based on a semantic match between the search terms and the at least one item keyword.

6. The system of claim 1 , wherein the search terms are determined by a keyword determination engine.

7. A computer-implemented method comprising:
generating, by a server and based on an exponentially time decayed algorithm, aggregated impression data by monitoring a frequency of display of each of a plurality of elements in one or more prior interfaces over a specified period of time;
generating, by the server and based on an exponentially time decayed algorithm, aggregated engagement data by monitoring a frequency of interaction of each of the plurality of elements in the one or more prior interfaces over a specified period of time;
receiving, from the server at a processor, a search request;
extracting search terms from the search request by applying a natural language processing algorithm to the search request;
identifying at least one search keyword by parsing the search terms;
determining a plurality of campaigns, wherein each of the plurality of campaigns includes data elements representative of at least one item keyword and at least one corresponding item, wherein the plurality of campaigns includes campaigns having at least one item keyword corresponding to the at least one search keyword;
determining an overall engagement probability for each of the plurality of campaigns, wherein the overall engagement probability is determined by:
generating an item engagement probability distribution by a posterior probability distribution of the aggregated impression data and the aggregated engagement data of the at least one corresponding item, wherein the posterior probability distribution is based on a binomial and a beta distribution;
generating, by a machine learning model and based on the item engagement probability distribution, an item engagement probability for the at least one corresponding item of each of the plurality of campaigns, wherein the item engagement probability is determined by a posterior probability distribution of the aggregated impression data and the aggregated engagement data; and
generating the overall engagement probability for each campaign by averaging the item engagement probability for each item of the campaign or selecting a maximum item engagement probability for the at least one corresponding item;

selecting one of the plurality of campaigns having a highest overall engagement probability;
providing output data to the server including the at least one corresponding item for the selected one of the plurality of campaigns;
generating, by the server, an interface including the at least one corresponding item;
providing, by the server, the interface including the at least one corresponding item to a device that generated the search request;
receiving impression data and engagement data for the interface including the at least one corresponding item associated with the selected one of the plur

Assignees

• Walmart Apollo Llc

Inventors

• Yang Peng

Classifications

• G06Q30/0256 Primary

  User search · CPC title

• G06F16/951

  Indexing; Web crawling techniques · CPC title

• G06N7/01

  Probabilistic graphical models, e.g. probabilistic networks · CPC title

• G06Q30/0277

  Online advertisement · CPC title

• G06N20/00

  Machine learning · CPC title