# US 2025/0094728 A1 · Summary of reviews generated by a generative language model

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

- Number: US 2025/0094728 A1
- Document ID: US20250094728A1
- Status: Published application
- Kind: A1
- Application number: 18/467,995
- Filed: 2023-09-15
- Published: 2025-03-20
- Earliest claimed: 2023-09-15
- Assignees: Shopify Inc.
- Applicants: Shopify Inc.
- Inventors: Chuanye Ouyang; Benjamin Lerchin; Danica Dillera
- Web page: https://patentgrep.com/patent/US20250094728A1

## CPC

- G06F 40/40: Processing or translation of natural language (natural language analysis G06F40/20; semantic analysis G06F40/30)
- G06F 16/345: {Summarisation for human users}
- G06F 40/166: Editing, e.g. inserting or deleting
- G06F 40/284: Lexical analysis, e.g. tokenisation or collocates
- G06F 40/56: Natural language generation
- G06N 20/00: Machine learning
- G06Q 30/0282: Rating or review of business operators or products
- G06Q 30/0631: {Recommending goods or services}
- G06F 40/30: Semantic analysis
- G06F: ELECTRIC DIGITAL DATA PROCESSING (computer systems based on specific computational models G06N)
- 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

A computer-implemented method is provided. The computer-implemented method may include: assigning relevancy values to reviews of a plurality of reviews; aggregating a subset of reviews from the plurality of reviews into an input prompt, the subset of reviews selected based on the relevancy values assigned to reviews in the plurality of reviews; and inputting the input prompt into a generative language model yielding a summary review of the subset of reviews generated by the generative language model.

## Claims

### Claim 1

A computer-implemented method comprising:
assigning relevancy values to reviews of a plurality of reviews;
aggregating a subset of reviews from the plurality of reviews into an input prompt, the subset of reviews selected based on the relevancy values assigned to reviews in the plurality of reviews; and
inputting the input prompt into a generative language model yielding a summary review of the subset of reviews generated by the generative language model.

### Claim 2 (depends on 1)

The computer-implemented method of claim 1, wherein the relevancy values assigned to a particular review of the plurality of reviews is based on at least one of informational density of review content of the particular review or keywords in the review content of the particular review.

### Claim 3 (depends on 1)

The computer-implemented method of claim 1, wherein the relevancy value assigned to a particular review of the plurality of reviews is based on review metadata associated with the particular review.

### Claim 4 (depends on 1)

The computer-implemented method of claim 1, wherein inputting the input prompt into the generative language model yielding the summary review comprises:
inputting the input prompt into the generative language model yielding attribute data associated with the subset of reviews generated by the generative language model; and
inputting the attribute data into the generative language model yielding an initial summary generated by the generative language model.

### Claim 5 (depends on 4)

The computer-implemented method of claim 4, further comprising:
inputting the initial summary and context including one or more example summary reviews into the generative language model yielding a final summary generated by the generative language model, wherein the final summary comprises the summary review.

### Claim 6 (depends on 1)

The computer-implemented method of claim 1, wherein the subset of reviews comprises reviews of the plurality of reviews having highest relevancy values.

### Claim 7 (depends on 1)

The computer-implemented method of claim 1, wherein the subset of reviews comprises a subset of reviews having a defined number of reviews, and the computer-implemented method further comprises determining the defined number of reviews by:
generating a plurality of subsets of reviews, wherein each subset of reviews of the plurality of subsets of reviews includes a number of reviews selected according to a review number sequence and wherein a subset of reviews in the review number sequence includes all reviews of a preceding subset of reviews in the review number sequence as well as additional reviews;
for each subset of reviews of the plurality of subsets of reviews:
generating the summary review of the subset of reviews using the generative language model, and
determining a distance between an embedding of the summary review of the subset of reviews and a corresponding embedding of a preceding summary review of the preceding subset of reviews; and
determining, as the defined number of reviews, a number of reviews in a particular subset of reviews of the plurality of subsets of reviews having the distance below a distance threshold.

### Claim 8 (depends on 7)

The computer-implemented method of claim 7, wherein the preceding subset of reviews comprises an immediately preceding subset of reviews in the review number sequence.

### Claim 9

A non-transitory computer-readable storage medium having stored thereon computer-executable instructions that, when executed, cause at least one processor to perform operations comprising:
assigning relevancy values to reviews of a plurality of reviews;
aggregating a subset of reviews from the plurality of reviews into an input prompt, the subset of reviews selected based on the relevancy values assigned to reviews in the plurality of reviews; and
inputting the input prompt into a generative language model yielding a summary review of the subset of reviews generated by the generative language model.

### Claim 10 (depends on 9)

The non-transitory computer-readable storage medium of claim 9, wherein the relevancy values assigned to a particular review of the plurality of reviews is based on at least one of:
informational density of review content of the particular review;
keywords in the review content of the particular review; or
review metadata associated with the particular review.

### Claim 11 (depends on 9)

The non-transitory computer-readable storage medium of claim 9, wherein the computer-executable instructions which cause the at least one processor to perform operations comprising inputting the input prompt into the generative language model yielding the summary review comprises instructions which cause the at least one processor to perform operations comprising:
inputting the input prompt into the generative language model yielding attribute data associated with the subset of reviews generated by the generative language model; and
inputting the attribute data into the generative language model yielding an initial summary generated by the generative language model.

### Claim 12 (depends on 11)

The non-transitory computer-readable storage medium of claim 11, further comprising computer-executable instructions which cause the at least one processor to perform operations comprising:
inputting the initial summary and context including one or more example summary reviews into the generative language model yielding a final summary generated by the generative language model, wherein the final summary comprises the summary review.

### Claim 13 (depends on 9)

The non-transitory computer-readable storage medium of claim 9, wherein the subset of reviews comprises reviews of the plurality of reviews having highest relevancy values.

### Claim 14 (depends on 9)

The non-transitory computer-readable storage medium of claim 9, wherein the subset of reviews comprises a subset of reviews having a defined number of reviews, and further comprising computer-executable instructions which cause the at least one processor to determine the defined number of reviews by performing operations comprising:
generating a plurality of subsets of reviews, wherein each subset of reviews of the plurality of subsets of reviews includes a number of reviews selected according to a review number sequence and wherein a subset of reviews in the review number sequence includes all reviews of a preceding subset of reviews in the review number sequence as well as additional reviews;
for each subset of reviews of the plurality of subsets of reviews:
generating the summary review of the subset of reviews using the generative language model, and
determining a distance between an embedding of the summary review of the subset of reviews and a corresponding embedding of a preceding summary review of the preceding subset of reviews; and
determining, as the defined number of reviews, a number of reviews in a particular subset of reviews of the plurality of subsets of reviews having the distance below a distance threshold.

### Claim 15 (depends on 14)

The non-transitory computer-readable storage medium of claim 14, wherein the preceding subset of reviews comprises an immediately preceding subset of reviews in the review number sequence.

### Claim 16

A system comprising:
at least one processor; and
a non-transitory computer-readable storage medium having stored thereon computer-executable instructions that, when executed, cause the at least one processor to:
assign relevancy values to reviews of a plurality of reviews;
aggregate a subset of reviews from the plurality of reviews into an input prompt, the subset of reviews selected based on the relevancy values assigned to reviews in the plurality of reviews; and
input the input prompt into a generative language model yielding a summary review of the subset of reviews generated by the generative language model.

### Claim 17 (depends on 16)

The system of claim 16, wherein the relevancy values assigned to a particular review of the plurality of reviews is based on at least one of:
informational density of review content of that review;
keywords in the review content of the particular review; or
review metadata associated with that review.

### Claim 18 (depends on 16)

The system of claim 16, wherein the computer-executable instructions which cause the at least one processor to perform operations comprising inputting the input prompt into the generative language model yielding the summary review comprises instructions which cause the at least one processor to:
input the input prompt into the generative language model yielding attribute data associated with the subset of reviews generated by the generative language model;
input the attribute data into the generative language model yielding an initial summary generated by the generative language model; and
input the initial summary and context including one or more example summary reviews into the generative language model yielding a final summary generated by the generative language model, wherein the final summary comprises the summary review.

### Claim 19 (depends on 16)

The system of claim 16, wherein the subset of reviews comprises reviews of the plurality of reviews having highest relevancy values.

### Claim 20 (depends on 16)

The system of claim 16, wherein the subset of reviews comprises a subset of reviews having a defined number of reviews, and the non-transitory computer-readable storage medium further stores computer-executable instructions which cause the at least one processor to determine the defined number of reviews by causing the at least one processor to:
generate a plurality of subsets of reviews, wherein each subset of reviews of the plurality of subsets of reviews includes a number of reviews selected according to a review number sequence and wherein a subset of reviews in the review number sequence includes all reviews of a preceding subset of reviews in the review number sequence as well as additional reviews;
for each subset of reviews of the plurality of subsets of reviews:
generate the summary review of the subset of reviews using the generative language model, and
determine a distance between an embedding of the summary review of the subset of reviews and a corresponding embedding of a preceding summary review of the preceding subset of reviews; and
determine, as the defined number of reviews, a number of reviews of a particular subset of reviews of the plurality of subsets of reviews having the distance below a distance threshold.
