# US 12,770,172 B2 · Systems and methods for automated training data generation for item attributes

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

- Number: US 12,770,172 B2
- Document ID: US12770172B2
- Status: Granted
- Kind: B2
- Application number: 17/516,089
- Filed: 2021-11-01
- Published: 2023-05-04
- Granted: 2026-10-20
- Earliest claimed: 2021-11-01
- Assignees: Walmart Apollo, LLC
- Applicants: Walmart Apollo, LLC
- Inventors: Adithya Rajan; Prateek Verma; Yilei Zhan; Nidhin Tom Pattaniyil; Zheng Yan
- Web page: https://patentgrep.com/patent/US12770172B2

## CPC

- G06Q 30/0627: {by specifying product or service characteristics, e.g. product dimensions}
- G06F 16/24575: {using context}
- G06F 16/907: Retrieval characterised by using metadata, e.g. metadata not derived from the content or metadata generated manually
- G06N 20/20: Ensemble learning
- G06Q 30/0201: Market modelling; Market analysis; Collecting market data
- G06Q 30/0204: Market segmentation
- G06Q 30/0603: {Catalogue creation or management}
- 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
- G06F: ELECTRIC DIGITAL DATA PROCESSING (computer systems based on specific computational models G06N)
- G06N: COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS

## Abstract

A data generation system can include a computing device that is configured to receive a request to generate a training dataset for an attribute and identify a set of item identifiers from an item database based on an engagement indication. The computing device is further configured to, for each item identifier of the set of item identifiers, obtain a query list including queries resulting in an engagement between the corresponding item identifier and a user and, in response to a portion of queries of the query list including the attribute being above a threshold, assign the corresponding item identifier to the training dataset for the attribute. The computing device is also configured to store the training dataset for the attribute in a training dataset database.

## Claims

### Claim 1

A system comprising:
a processor; and
a non-transitory memory storing instructions that, when executed, cause the processor to:
receive a request to generate a training dataset for an attribute defining one or more features of an item in a first plurality of items, wherein the attribute and an attribute threshold value are defined, at least in part, by the request;
determine a similarity score for each item identifier, in a plurality of item identifiers, associated with a corresponding plurality of items, different from the first plurality of items, stored in an item database based on one or more queries, each respective query in the one or more queries is (i) associated, at least in part, with one or more corresponding interactions with a respective item, in a first subset of a plurality of items, displayed at a corresponding computing device and (ii) received, at least in part, as one or more words defining the attribute and inputted at a graphical user interface of the corresponding computing device, in a plurality of computing devices;
select a first set of item identifiers, in the plurality of item identifiers, from the item database based on the similarity score, wherein the similarity score indicates how related a corresponding item identifier is to the attribute;
generate the training dataset for the attribute by:
for each item identifier of the first set of item identifiers:
compare the similarity score to the attribute threshold value; and
in response to the similarity score being above the attribute threshold value, include the corresponding item identifier in the training dataset for the attribute;
store the training dataset for the attribute in a training dataset database different from the item database;
generate a trained machine learning model using the training dataset for the attribute applied to a machine learning model;
store the trained machine learning model in a model database different from the item database and the training dataset database;
receive a request to display a first item, different from the first plurality of items and the corresponding plurality of items, via a corresponding graphical user interface of a respective computing device in the plurality of computing devices, the request comprising a second set of item identifiers, different from the first set of item identifiers, associated with the first item; and
use the trained machine learning model to perform a process comprising:
select a subsequent set of item identifiers, in the plurality of item identifiers, from the item database based on the first item and a subsequent similarity score generated by the trained machine learning model using the second set of item identifiers, wherein the subsequent similarity score comprises a value defining a relation between an item identifier in the subsequent set of item identifiers is to the attribute;
receive an updated attribute threshold value;
for each item identifier of the subsequent set of item identifiers:
compare the similarity score to the updated attribute threshold value; and
in response to the similarity score being above the attribute threshold value, assign the corresponding item identifier to an updated training dataset for the attribute;
store (i) the updated training dataset for the attribute in the training dataset database and (ii) the corresponding item identifier associated with the first item in the item database; and
generate, responsive to receiving an input comprising a first set of words, different from the one or more words, defining the attribute and inputted at a graphical user interface of a first computing device, one or more instructions causing display at the first computing device of at least the first item.

### Claim 2 (depends on 1)

The system of claim 1, wherein the instructions further cause the processor to:
identify the set of item identifiers from the item database based on the similarity score by selecting item identifiers from the item database including a corresponding order frequency above an order threshold.

### Claim 3 (depends on 1)

The system of claim 1, wherein the instructions further cause the processor to:
identify the set of item identifiers from the item database based on the similarity score by:
determining an engagement value based on at least one of a number of orders, a number of add-to-cart selections, and a number of view selections; and
selecting a predetermined number of item identifiers corresponding to highest engagement values.

### Claim 4 (depends on 1)

The system of claim 1, wherein the instructions further cause the processor to:
identify the set of item identifiers from the item database based on the similarity score by:
determining an engagement value based on at least one of a number of orders, a number of add-to-cart selections, and a number of view selections; and
selecting the set of item identifiers as item identifiers with a corresponding engagement value above a first engagement threshold.

### Claim 5 (depends on 1)

The system of claim 1, wherein comparing the similarity score to the attribute threshold value includes identifying a subset of queries of a query list including a number of engagements between the corresponding item identifier and a user being above a second engagement threshold.

### Claim 6 (depends on 1)

The system of claim 1, wherein the attribute includes at least one of: (i) a gender, (ii) an age, and (iii) a color.

### Claim 7 (depends on 1)

The system of claim 1, wherein the instructions further cause the processor to:
classify item identifiers as including the attribute using the trained machine learning model.

### Claim 8 (depends on 1)

The system of claim 1, wherein the instructions further cause the processor to, in response to receiving a new item identifier:
determine at least one attribute of the new item identifier by applying a plurality of machine learning models, comprising the trained machine learning model, stored in the model database to the new item identifier; and
identify and tag the new item identifier based on the at least one attribute.

### Claim 9

A method comprising:
receiving a request to generate a training dataset for an attribute defining one or more features of an item in a first plurality of items, wherein the attribute and an attribute threshold value are defined, at least in part, by the request;
determining a similarity score for each item identifier, in a plurality of item identifiers, associated with a corresponding plurality of items, different from the first plurality of items, stored in an item database based on one or more queries, each respective query in the one or more queries is (i) associated, at least in part, with one or more corresponding interactions with a respective item, in a first subset of a plurality of items, displayed at a corresponding computing device and (ii) received, at least in part, as one or more words defining the attribute and inputted at a graphical user interface of the corresponding computing device, in a plurality of computing devices;
selecting a first set of item identifiers, in the plurality of item identifiers, from the item database based on the similarity score, wherein the similarity score indicates how related a corresponding item identifier is to the attribute;
generating the training dataset for the attribute by:
for each item identifier of the first set of item identifiers:
comparing the similarity score to the attribute threshold value; and
in response to the similarity score being above the attribute threshold value, including the corresponding item identifier in the training dataset for the attribute;
storing the training dataset for the attribute in a training dataset database different from the item database;
generating a trained machine learning model using the training dataset for the attribute applied to a machine learning model;
storing the trained machine learning model in a model database different from the item database and the training dataset database;
receiving a request to display a first item, different from the first plurality of items and the corresponding plurality of items, via a corresponding graphical user interface of a respective computing device in the plurality of computing devices, the request comprising a second set of item identifiers, different from the first set of item identifiers, associated with the first item; and
using the trained machine learning model to perform a process comprising:
selecting a subsequent set of item identifiers, in the plurality of item identifiers, from the item database based on the first item and a subsequent similarity score generated by the trained machine learning model using the second set of item identifiers, wherein the subsequent similarity score comprises a value defining a relation between an item identifier in the subsequent set of item identifiers is to the attribute;
receiving an updated attribute threshold value;
for each item identifier of the subsequent set of item identifiers:
comparing the similarity score to the updated attribute threshold value; and
in response to the similarity score being above the attribute threshold value, assign the corresponding item identifier to an updated training dataset for the attribute;
storing (i) the updated training dataset for the attribute in the training dataset database and (ii) the corresponding item identifier associated with the first item in the item database; and
generating, responsive to receiving an input comprising a first set of words, different from the one or more words, defining the attribute and inputted at a graphical user interface of a first computing device, one or more instructions causing display at the first computing device of at least the first item.

### Claim 10

A non-transitory computer readable medium having instructions stored thereon, wherein the instructions, when executed by at least one processor, cause a device to perform operations comprising:
receiving a request to generate a training dataset for an attribute defining one or more features of an item in a first plurality of items, wherein the attribute and an attribute threshold value are defined, at least in part, by the request;
determining a similarity score for each item identifier, in a plurality of item identifiers, associated with a corresponding plurality of items, different from the first plurality of items, stored in an item database based on one or more queries, each respective query in the one or more queries is (i) associated, at least in part, with one or more corresponding interactions with a respective item, in a first subset of a plurality of items, displayed at a corresponding computing device and (ii) received, at least in part, as one or more words defining the attribute and inputted at a graphical user interface of the corresponding computing device, in a plurality of computing devices;
selecting a first set of item identifiers, in the plurality of item identifiers, from the item database based on the similarity score, wherein the similarity score indicates how related a corresponding item identifier is to the attribute;
generating the training dataset for the attribute by:
for each item identifier of the first set of item identifiers:
comparing the similarity score to the attribute threshold value; and
in response to the similarity score being above the attribute threshold value, including the corresponding item identifier in the training dataset for the attribute;
storing the training dataset for the attribute in a training dataset database different from the item database;
generating a trained machine learning model using the training dataset for the attribute applied to a machine learning model;
storing the trained machine learning model in a model database different from the item database and the training dataset database;
receiving a request to display a first item, different from the first plurality of items and the corresponding plurality of items, via a corresponding graphical user interface of a respective computing device in the plurality of computing devices, the request comprising a second set of item identifiers, different from the first set of item identifiers, associated with the first item; and
using the trained machine learning model to perform a process comprising:
selecting a subsequent set of item identifiers, in the plurality of item identifiers, from the item database based on the first item and a subsequent similarity score generated by the trained machine learning model using the second set of item identifiers, wherein the subsequent similarity score comprises a value defining a relation between an item identifier in the subsequent set of item identifiers is to the attribute;
receiving an updated attribute threshold value;
for each item identifier of the subsequent set of item identifiers:
comparing the similarity score to the updated attribute threshold value; and
in response to the similarity score being above the attribute threshold value, assigning the corresponding item identifier to an updated training dataset for the attribute;
storing (i) the updated training dataset for the attribute in the training dataset database and (ii) the corresponding item identifier associated with the first item in the item database; and
generating, responsive to receiving an input comprising a first set of words, different from the one or more words, defining the attribute and inputted at a graphical user interface of a first computing device, one or more instructions causing display at the first computing device of at least the first item.

### Claim 11 (depends on 1)

The system of claim 1, wherein the determining the similarity score is based, at least in part, on a first weight associated with a first type of interaction in the one or more corresponding interactions and a second weight, different from the first weight, associated with a second type of interaction in the one or more corresponding interactions, different from the first type of interaction.

### Claim 12 (depends on 1)

The system of claim 1, wherein the determining the similarity score comprises assigning a first value to each respective interaction in the one or more corresponding interactions that satisfies a threshold value and assigning a second value, less than the first value, that fails to satisfy the threshold value.

### Claim 13 (depends on 1)

The system of claim 1, wherein the attribute is based, at least in part, on a text description of a first length and a text description of a second length greater than the first length and defined, at least in part, by a third-party.

### Claim 14 (depends on 1)

The system of claim 1, wherein the set of items is selected based, at least in part, on a predetermined threshold number of items associated with the set of items.

### Claim 15 (depends on 1)

The system of claim 1, wherein the similarity score is bound between a minimum respective similar score and a maximum respective similarity score.

### Claim 16 (depends on 1)

The system of claim 1, wherein the one or more corresponding interactions comprises a viewing of a respective item through a graphical user interface, a selection of the respective item through the graphical user interface, a value of the respective item, or a combination thereof.

### Claim 17 (depends on 1)

The system of claim 1, wherein, prior to the determining the similarity score comprises, the instructions further cause the processor to identify each item stored in the item database satisfying a threshold value of queries.

### Claim 18 (depends on 1)

The system of claim 1, wherein the attribute includes (i) a gender, (ii) an age, (iii) a color, or (iv) a size.

### Claim 19 (depends on 1)

The system of claim 1, wherein the training dataset comprises a corresponding item definition for each respective item in the set of item identifiers.

### Claim 20 (depends on 1)

The system of claim 1, wherein the updated training dataset is accessible to a plurality of untrained models.
