Automated Item Correlation via Multi-Modal Confidence Scoring

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Conventional methods for correlating item data in competitive analysis are manual, error-prone, and result in obsolete connections, leading to misinformation and inefficient resource usage due to incorrect item associations.

Innovation Solution

A system that receives reference text and image data, determines similarity scores using text and image correlation models, calculates confidence scores, and performs responsive actions such as creating associations or modifying models based on these scores to improve item correlation accuracy and resource management.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Reliability

If manual user searches and comparisons are used to correlate item data, then users can perform competitive analysis, but the process is slow, error-prone, and produces obsolete connections that misinform decision makers

Engineering Contradiction:

Improveaccuracy of item associationsVSAvoidtime to produce item correlations

Core Design Contradiction:

ReliabilityVSLoss of time

Solution Approach 1:

The patent replaces manual mechanical comparison processes with automated image recognition and text analysis systems. The system uses machine learning models to automatically compare product images, extract text data, and determine item similarities, eliminating the need for manual user searches and comparisons while significantly improving both speed and accuracy of item correlations

Inventive Principle:

Principle #28Mechanics substitution (Replace mechanical system)

Solution Approach 2:

The system enables self-service automated item correlation by using confidence scores to automatically determine whether items are associated or differentiated. The system self-corrects by learning from user feedback and automatically adjusting its correlation decisions, reducing reliance on manual intervention while maintaining high accuracy

Inventive Principle:

Principle #25Self-service

2Reliability

If users manually differentiate between items, then they can avoid incorrect associations, but this requires additional user intervention and system resources

Engineering Contradiction:

Improveaccuracy of item differentiationVSAvoidsystem resources for manual intervention

Core Design Contradiction:

ReliabilityVSDevice complexity

Solution Approach 1:

The system implements feedback mechanisms where user corrections of automated decisions are captured and used to retrain and improve the machine learning models. This feedback loop allows the system to learn from mistakes and continuously improve its differentiation accuracy, reducing the need for ongoing manual intervention while maintaining high reliability

Inventive Principle:

Principle #23Feedback

Solution Approach 2:

The patent introduces confidence scores as an intermediary metric that bridges automated analysis and final correlation decisions. The system calculates confidence scores based on image and text similarity, and uses these scores to automatically determine whether items are associated or differentiated, reducing the need for direct manual intervention while maintaining accurate differentiation

Inventive Principle:

Principle #24Intermediary (Mediator)

3Productivity

If automated systems create item associations quickly, then productivity increases, but misinformation and incorrect associations occur

Engineering Contradiction:

Improvespeed of item correlationVSAvoidmisinformation from incorrect associations

Core Design Contradiction:

ProductivityVSLoss of information

Solution Approach 1:

The patent replaces error-prone manual correlation processes with automated image recognition and text analysis systems that process multiple data points simultaneously. The system uses confidence scores derived from multiple features (image similarity, text matching, product specifications) to make informed correlation decisions, achieving high productivity while minimizing misinformation through multi-factor verification

Inventive Principle:

Principle #28Mechanics substitution (Replace mechanical system)

Solution Approach 2:

The system uses confidence scores as an intermediary validation mechanism between automated analysis and final associations. By calculating and evaluating confidence scores before creating associations, the system ensures that only high-quality, well-supported correlations are created, preventing misinformation while maintaining rapid automated processing

Inventive Principle:

Principle #24Intermediary (Mediator)

Data Source

PatentUS11709997B2Systems and methods for modeling item similarity and correlating item information

Publication Date: 2023.07.25 COUPANG CORP

AI summary

Disclosed herein are systems and methods for correlating item data. A system for correlating item data may comprise a memory storing instructions and at least one processor configured to execute instructions to perform operations comprising: receiving reference text data associated with a reference item from a device; receiving reference image data associated with the reference item from the remote device; determining candidate text data and candidate image data associated with at least one candidate item; selecting a text correlation model; determining a first similarity score by applying the text correlation model to the reference text data and the candidate text data; selecting an image correlation model; determining a second similarity score by applying the image correlation model to the reference image data and the candidate image data; calculating a confidence score based on the first and second similarity scores; and performing a responsive action based on the calculated confidence score.