Review Summarization Prompting with Relevancy-Based Selection

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Existing generative language models face limitations due to input prompt limits and increased processing time and resources when summarizing large numbers of reviews, leading to lower quality summaries when irrelevant or uninformative reviews are included.

Innovation Solution

A method that assigns relevancy values to reviews, selects a subset based on these values, and uses a generative language model to generate summaries, incorporating standardized input prompts to improve quality and efficiency.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Loss of information

If all reviews are aggregated into the input prompt for summarization, then the completeness of information is improved, but the processing time and computing resources increase significantly

Engineering Contradiction:

Improvecompleteness of informationVSAvoidprocessing time

Core Design Contradiction:

Loss of informationVSLoss of time

Solution Approach 1:

The patent extracts only the most relevant reviews from the complete set of reviews based on relevancy scoring. This selective extraction allows the system to maintain information quality while reducing the volume of data processed, thereby decreasing processing time and computing resource consumption without losing critical information.

Inventive Principle:

Principle #2Taking out (Extraction)

Solution Approach 2:

The patent segments the complete set of reviews into relevant and irrelevant portions based on relevancy criteria. By processing only the segmented relevant subset rather than the entire dataset, the system achieves efficient processing while preserving the essential information needed for accurate summarization.

Inventive Principle:

Principle #1Segmentation

2Loss of information

If all reviews are aggregated into the input prompt for summarization, then the completeness of information is improved, but the quality of summary decreases due to inclusion of irrelevant reviews

Engineering Contradiction:

Improvecompleteness of informationVSAvoidquality of summary

Core Design Contradiction:

Loss of informationVSManufacturing precision

Solution Approach 1:

The patent extracts only the most relevant reviews from the complete set based on relevancy scoring. This selective extraction removes irrelevant or low-quality reviews that would degrade summary quality, while preserving the essential information needed for accurate and high-quality summarization.

Inventive Principle:

Principle #2Taking out (Extraction)

Solution Approach 2:

The patent applies different quality standards to different reviews by evaluating each review's relevancy. Reviews are selectively included based on their individual quality characteristics, ensuring that only high-quality, relevant information contributes to the final summary, thereby maintaining high manufacturing precision.

Inventive Principle:

Principle #3Local quality

3Loss of information

If a large number of reviews are aggregated into the input prompt, then the completeness of information is improved, but the input prompt exceeds the model's token limit

Engineering Contradiction:

Improvecompleteness of informationVSAvoidinput prompt limit

Core Design Contradiction:

Loss of informationVSAdaptability or versatility

Solution Approach 1:

The patent extracts only the necessary subset of reviews that fit within the model's token limit while preserving the most important information. This selective extraction ensures the input prompt remains within technical constraints while maintaining informational completeness for effective summarization.

Inventive Principle:

Principle #2Taking out (Extraction)

Solution Approach 2:

The patent processes only the necessary portion of reviews (partial action) rather than attempting to include all reviews. By processing a carefully selected subset that fits within token limits, the system achieves effective summarization without exceeding model constraints.

Inventive Principle:

Principle #16Partial or excessive action

4Loss of information

If all reviews are processed by the LLM, then the comprehensiveness of analysis is improved, but the computing resources required increase

Engineering Contradiction:

Improvecomprehensiveness of analysisVSAvoidcomputing resources

Core Design Contradiction:

Loss of informationVSUse of energy by moving object

Solution Approach 1:

The patent extracts only the most relevant reviews for LLM processing based on relevancy scoring. This selective extraction reduces the volume of data requiring computational processing while maintaining the comprehensiveness of analysis by ensuring all critical information is included in the processed subset.

Inventive Principle:

Principle #2Taking out (Extraction)

Solution Approach 2:

The patent segments reviews into relevant and irrelevant portions, processing only the relevant segment through the computationally intensive LLM. This segmentation approach maintains analytical comprehensiveness while significantly reducing computing resource consumption by excluding irrelevant reviews from processing.

Inventive Principle:

Principle #1Segmentation

Data Source

PatentUS12579377B2Summary of reviews generated by a generative language model

Publication Date: 2026.03.17 SHOPIFY INC

AI summary

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.