# US 11,443,180 B1 · Mapping content to an item repository

> Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt

- Number: US 11,443,180 B1
- Document ID: US11443180B1
- Status: Granted
- Kind: B1
- Application number: 15/994,274
- Filed: 2018-05-31
- Granted: 2022-09-13
- Earliest claimed: 2018-05-31
- Assignees: A9.com, Inc.
- Applicants: A9.com, Inc.
- Inventors: Adrian Boteanu; Shay Artzi; Adam Kiezun; Raju Matta
- Web page: https://patentgrep.com/patent/US11443180B1

## CPC

- G06N 3/0455: Auto-encoder networks; Encoder-decoder networks
- G06N 3/044: Recurrent networks, e.g. Hopfield networks
- G06N 3/0442: characterised by memory or gating, e.g. long short-term memory [LSTM] or gated recurrent units [GRU]
- G06N 3/045: Combinations of networks
- G06N 3/0464: Convolutional networks [CNN, ConvNet]
- G06N 3/08: Learning methods
- G06N 3/082: modifying the architecture, e.g. adding, deleting or silencing nodes or connections
- G06N 3/0895: Weakly supervised learning, e.g. semi-supervised or self-supervised learning
- G06N 3/09: Supervised learning
- G06N 3/096: Transfer learning
- G06Q 30/0631: {Recommending goods or services}
- G06Q 30/0282: Rating or review of business operators or products
- G06N: COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06Q: INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR

## Abstract

Various embodiments of systems and methods for mapping review articles to locations in a marketplace catalog. Various embodiments include training a system using product reviews and images from the marketplace as training data. Product reviews and images can be provided to autoencoders to generate respective short codes. The product review short codes can be combined with the product image short codes to generate review short codes. These review short codes can be provided as training data for an attribute assignment network while associated attributes (e.g., attributes associated with related products) can be target outputs for the attribute assignment network. After the system is sufficiently trained, a review article can be obtained from the publishing platform. This review article can be provided to the trained system to receive predicted attribute assignments. These predicted attribute assignments can be effective to map the review article to locations in a marketplace catalog.

## Claims

### Claim 1

A computer-implemented method for training a location assignment system, comprising:
obtaining, from an entity that provides a plurality of products, training content including at least product reviews and product images associated with the plurality of products;
generating product review short codes by supplying the product reviews to a text autoencoder;
generating product image short codes by supplying the product images to the image autoencoder;
combining, for individual products of the plurality of products, the product review short codes and the product image short codes to form training short codes;
training an assignment network using the training short codes as inputs and related marketplace item repository locations as target outputs, the related marketplace item repository locations associated with at least one of the plurality of products;
obtaining articles from a review publisher, the articles having article text portions and article images;
determining one or more locations in an item repository for the articles using the text autoencoder, the image autoencoder, and the assignment network; and
storing a mapping of the one or more locations to the articles from which the articles are accessible.

### Claim 2 (depends on 1)

The computer-implemented method of claim 1, further comprising:
receiving a request at the marketplace for information associated with an item repository location of the one or more locations;
determine, using on the mapping, that the item repository location maps to an article; and
return information for the article.

### Claim 3 (depends on 1)

The computer-implemented method of claim 1, wherein the text autoencoder is a stacked long short-term memory (LSTM) autoencoder with a sentence autoencoding layer, a paragraph autoencoding layer, and a review text autoencoding layer.

### Claim 4

A computer-implemented method, comprising:
obtaining a training text and a training image, the training text and the training image being associated with a location in an item repository;
generating, from the training text and using a text autoencoder, a training text short code;
generating, from the training image and using an image autoencoder, a training image short code;
generating a training short code from the training text short code and the training short code;
training an assignment network using at least the training short code as an input and the location as a target output, yielding a trained assignment network;
obtaining an additional text and an additional image, the additional text and the additional image being associated with a content item;
processing the additional text and the additional image using the text autoencoder, the image autoencoder, and the trained assignment network to determine an additional location in the item repository; and
associate the additional location to the content item.

### Claim 5 (depends on 4)

The computer-implemented method of claim 4, further comprising:
receiving a request for information associated with the additional location in the item repository; and
providing, in response to the request, information for the content item.

### Claim 6 (depends on 5)

The computer-implemented method of claim 5, wherein the training text, the training image, and item repository are associated with a first domain and the content item is associated with a second domain.

### Claim 7 (depends on 4)

The computer-implemented method of claim 4, further comprising:
receiving a request for the content item; and
providing, in response to the request, a reference to the additional location in the item repository.

### Claim 8 (depends on 4)

The computer-implemented method of claim 4, wherein generating a training short code comprises using a combination autoencoder.

### Claim 9 (depends on 4)

The computer-implemented method of claim 4, further comprising:
processing a content item short code for the content item using a second trained assignment network for a second repository to determine a second additional location on the second repository; and
associating the second additional location with the content item.

### Claim 10 (depends on 4)

The computer-implemented method of claim 4, further comprising:
obtaining a second content item;
determining that the second content item comprises a first portion and a second portion, the first portion having a first portion text and a first portion image, and the second article portion having a second portion text and a second portion image;
processing the first portion text and the first portion image using the text autoencoder, the image autoencoder, and the trained assignment network to determine a first portion location in the item repository;
processing the second portion text and the second portion image using the text autoencoder, the image autoencoder, and the trained assignment network to determine a second portion location in the item repository; and
associating the first portion location and the second portion location with the second content item.

### Claim 11 (depends on 4)

The computer-implemented method of claim 4, wherein the text autoencoder includes a text convolutional autoencoder or a stacked long short-term memory (LSTM) autoencoder with a sentence autoencoding layer, a paragraph autoencoding layer, and a full text autoencoding layer.

### Claim 12 (depends on 4)

The computer-implemented method of claim 4, wherein the image autoencoder is a convolutional autoencoder, the method further comprising resizing the training image to create a resized training image and providing the resized product image as an input to the image autoencoder.

### Claim 13 (depends on 4)

The computer-implemented method of claim 4, further comprising:
determining that the location for the item repository needs more training data;
selecting a second training text and a second training image, wherein the second training text and the second training image are for different items that are associated with the location; and
further training the assignment network using a second training short code generated from the second training text and the second training image.

### Claim 14

A system, comprising:
at least one processor; and
memory including instructions that, when executed by the at least one processor, cause the system to:
obtain a training text and a training image, the training text and the training image being associated with one or more labels of an item repository;
generate, from the training text using a text autoencoder, a training text short code;
generate, from the training image using an image autoencoder, a training image short code;
generate a training short code from the training text short code and the training image short code;
train an assignment network using at least the training short code as an input and the one or more labels of a catalog as a target output, yielding a trained assignment network;
obtain an additional text and an additional image, the additional text and the additional image being associated with a content item;
processing the additional text and the additional image using the text autoencoder, the image autoencoder, and the trained assignment network to determine at least one label of the item repository; and
associate the at least one label to the content item.

### Claim 15 (depends on 14)

The system of claim 14, wherein the instructions when executed further cause the system to:
receive a request for information associated with the at least one label of the item repository; and
provide, in response to the request, information for the content item.

### Claim 16 (depends on 15)

The system of claim 15, wherein the training text, the training image, and the item repository are associated with a first domain and the content item is associated with a second domain.

### Claim 17 (depends on 14)

The system of claim 14, wherein the instructions when executed further cause the system to:
receive a request for information associated with the content item; and
provide, in response to the request, information associated with the at least one label of the item repository.

### Claim 18 (depends on 14)

The system of claim 14, wherein the instructions that, when executed, cause the system to generate a training short code, include instructions to use the combination autoencoder.

### Claim 19 (depends on 14)

The system of claim 14, wherein the text autoencoder is a stacked long short-term memory (LSTM) autoencoder with a sentence autoencoding layer, a paragraph autoencoding layer, and a full text autoencoding layer.

### Claim 20 (depends on 14)

The system of claim 14, wherein the instructions when executed further cause the system to:
determine that the at least one label of the item repository needs more training data;
selecting a second training text and a second training image, wherein the second training text and the second training image are for different items that are associated with the at least one label; and
training the assignment network using a second training short code generated from the second training review text and the second training image.