Artificial intelligence system for automated selection and presentation of informational content

Patent metadata

Field	Value
Publication number	US-10909604-B1
Application number	US-201815914923-A
Country	US
Kind code	B1
Filing date	Mar 7, 2018
Priority date	Mar 7, 2018
Publication date	Feb 2, 2021
Grant date	Feb 2, 2021

Abstract

A set of informational content elements pertaining to an item for presentation to one or more potential item consumers is identified at an artificial intelligence service. A plurality of optimization iterations are implemented. In a particular iteration, a set of content elements to be presented to a target audience in accordance with a set of presentation constraints indicated by a content source associated with the item is identified using a machine learning model, and metrics indicating the effectiveness of the content elements are analyzed.

First claim

What is claimed is:

1. A system, comprising:
one or more computing devices of a network-based content provider configured to present informational content for a plurality of items in one or more presentation contexts over a network for a plurality of sources of the items;
one or more other computing devices of an artificial intelligence service for optimizing presentation of informational content for at least some of the plurality of items;
wherein the one or more other computing devices of the artificial intelligence service are configured to:
classify the plurality of items to identify a subset of the items as candidate items for optimized selection of informational content elements;
send respective notifications of the candidate items, via one or more programmatic interfaces, to respective entities authorized to provide informational content about the candidate items, and request that the respective entities provide informational content elements for the candidate items for optimized selection of informational content elements;
receive from individual ones of the entities for individual ones of the candidate items, via the one or more programmatic interfaces, (a) a plurality of informational content elements pertaining to individual candidate items and (b) one or more presentation constraints associated with the individual candidate items, wherein the plurality of informational content elements comprises at least a first example and a second example of a particular content element category of a plurality of content element categories, and wherein a particular constraint of the one or more presentation constraints comprises a content portion inclusion constraint;
generate for individual ones of the candidate items, corresponding to individual ones of the plurality of informational content elements, a respective set of input features for a machine learning model; and
perform for individual ones of the candidate items a plurality of optimization iterations using the machine learning model, wherein input to the machine learning model comprises one or more of the respective sets of input features, and wherein a particular optimization iteration of the plurality of optimization iterations comprises:
determining, from the plurality of informational content elements in accordance with the one or more presentation constraints, a particular set of informational content elements to be presented to a first target audience in a first presentation context of the one or more presentation contexts, wherein the particular set comprises the first example, and wherein the particular set does not comprise the second example;
providing the particular set of informational content elements to the network-based content provider for subsequent transmission to the first target audience in the first presentation context;
obtaining from the network-based content provider one or more metrics associated with one or more interactions by the first target audience with a presentation of the particular set of informational content elements to the first target audience in the first presentation context; and
adjusting, based at least in part on an analysis of the obtained one or more metrics, the particular set of informational content elements to be presented by the network-based content provider in a future iteration; and

wherein the one or more computing devices of the network-based content provider are configured to:
present the particular set of informational content elements to the first target audience during the particular optimization iteration; and
detect the one or more interactions by the first target audience with the presentation of the particular set of informational content elements.

2. The system as recited in claim 1 , wherein the particular set of informational content elements comprises a modified version of a first informational content element obtained from the source, wherein the modified version is generated at the artificial intelligence service using one or more additional machine learning models.

3. The system as recited in claim 1 , wherein the source associated with the particular item comprises a producer or author of the particular item.

4. The system as recited in claim 1 , wherein the machine learning model comprises a bandit model.

5. The system as recited in claim 1 , wherein the particular set of informational content elements comprises informational content elements of the plurality of content element categories including a first category and a second category, wherein the one or more computing devices are configured to:
determine, after a second particular optimization iteration of the plurality of optimization iterations, that (a) further exploration of examples of the first category is not to be performed and (b) further exploration of examples of the second category is to be performed;
include, in a set of informational content elements presented in one optimization iteration after the second particular optimization iteration, the first example of the first category and a first example of the second category; and
include, in a set of informational content elements presented in another optimization iteration after the second particular optimization iteration, the first example of the first category and a second example of the second category.

6. A method, comprising:
performing, by one or more computing devices of an artificial intelligence service:
classifying a plurality of items to identify a subset of the items as candidate items for optimized selection of informational content elements;
sending respective notifications of the candidate items to respective entities authorized to provide informational content about the candidate items, and requesting that the respective entities provide informational content elements for the candidate items for optimized presentation of informational content elements;
receiving from individual ones of the entities for individual ones of the candidate items (a) a plurality of informational content elements pertaining to individual candidate items and (b) one or more presentation constraints associated with the individual candidate items, wherein a particular constraint of the one or more presentation constraints comprises a content portion inclusion constraint; and
performing for individual ones of the candidate items a plurality of optimization iterations using a machine learning model, wherein a particular optimization iteration of the plurality of optimization iterations comprises:
determining, from the plurality of informational content elements in accordance with the one or more presentation constraints, a particular set of informational content elements to be presented to a first target audience in a first presentation context;
providing the particular set of informational content elements to a network-based content provider for subsequent transmission to the first target audience in the first presentation context;
obtaining from the network-based content provider one or more metrics associated with one or more interactions by the first target audience with a presentation of the particular set of informational content elements to the first target audience in the first presentation context; and
adjusting, based at least in part on an analysis of the obtained one or more metrics, the particular set of informational content elements to be presented by the network-based content provider in a future iteration; and

performing, by one or more other computing devices of the network-based content provider:
presenting the particular set of informational content elements to the first target audience in the first presentation context during the particular optimization iteration; and
detecting the one or more interactions by the firs

Assignees

• Amazon Tech Inc

Inventors

• Zappella Giovanni

Classifications

• G06Q30/0631 Primary

  Recommending goods or services · CPC title

• G06Q30/0623 Primary

  by investigating goods or services · CPC title

• G06F18/2415

  based on parametric or probabilistic models, e.g. based on likelihood ratio or false acceptance rate versus a false rejection rate · CPC title

• G06N3/092

  Reinforcement learning · CPC title

• G06N3/08

  Learning methods · CPC title