Neural Network IPQ Comparison for Item Description Matching

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Current machine learning models are not well adapted to detect item package quantity information in item descriptions, often failing to accurately recognize IPQ indicators due to their focus on other aspects, leading to inefficiencies in product matching and navigation within large electronic catalogs.

Innovation Solution

A machine learned neural network-based IPQ comparison system that utilizes token-level attributes to analyze item descriptions, tokenizing and encoding them to generate fixed-size feature vectors for precise IPQ comparison, enabling accurate detection of matching or differing quantities.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Ease of manufacture

If rule-based solutions are used to detect item package quantity information, then the system is simple to implement, but the applicability and recall are inadequate

Engineering Contradiction:

Improveease of implementationVSAvoidapplicability and recall

Core Design Contradiction:

Ease of manufactureVSReliability

Solution Approach 1:

The patent replaces rule-based mechanical text processing with a machine learning model that automatically learns patterns from training data. The model substitutes manual rule creation with automated pattern recognition, achieving both high applicability and ease of implementation through training on labeled examples rather than manual rule programming.

Inventive Principle:

Principle #28Mechanics substitution (Replace mechanical system)

Solution Approach 2:

The patent changes the approach from fixed rules to learned parameters. The machine learning model adjusts its internal parameters through training to optimize detection accuracy, allowing it to adapt to different item description formats and languages without requiring manual rule updates for each scenario.

Inventive Principle:

Principle #35Parameter changes

2Adaptability or versatility

If machine learning models focus on general aspects of item descriptions, then they can process diverse data, but they fail to accurately detect IPQ indicators

Engineering Contradiction:

Improveability to process diverse dataVSAvoidaccuracy of IPQ detection

Core Design Contradiction:

Adaptability or versatilityVSMeasurement precision

Solution Approach 1:

The patent applies local quality by training the model to focus on specific local patterns within item descriptions that indicate package quantity. Rather than treating all text equally, the model learns to identify and weight specific phrases, numbers, and contextual patterns that are locally relevant to IPQ detection while maintaining overall versatility.

Inventive Principle:

Principle #3Local quality

Solution Approach 2:

The patent segments the item description processing into distinct features that the model evaluates separately. By breaking down the text into tokens and analyzing specific linguistic patterns, the model can accurately detect IPQ indicators while maintaining the ability to process diverse description formats through its comprehensive feature set.

Inventive Principle:

Principle #1Segmentation

3Measurement precision

If human analysis is used to extract item quantity information, then high accuracy is achieved, but the cost becomes prohibitive at scale

Engineering Contradiction:

Improveaccuracy of quantity extractionVSAvoidcost efficiency at scale

Core Design Contradiction:

Measurement precisionVSProductivity

Solution Approach 1:

The patent implements self-service by creating a machine learning system that automatically extracts and analyzes item package quantity information without human intervention. The model trains on labeled data and then autonomously processes item descriptions, eliminating the need for costly human analysis while maintaining high accuracy through automated pattern recognition.

Inventive Principle:

Principle #25Self-service

4Productivity

If current machine learning models are used for item analysis, then they can handle large volumes of data, but they are not well adapted to detect IPQ information

Engineering Contradiction:

Improveability to handle large volumes of dataVSAvoidsuitability for IPQ detection

Core Design Contradiction:

ProductivityVSReliability

Solution Approach 1:

The patent applies preliminary action by pre-training the machine learning model on labeled item description data that includes IPQ indicators. This preliminary training equips the model with the specific knowledge needed to detect package quantity information, allowing it to then efficiently process large volumes of diverse item descriptions with high reliability.

Inventive Principle:

Principle #10Preliminary action

Data Source

PatentUS11461829B1Machine learned system for predicting item package quantity relationship between item descriptions

Publication Date: 2022.10.04 AMAZON TECH INC

AI summary

Systems and methods are disclosed to implement a machine learned system to determine the comparative relationship between item package quantity (IPQ) information indicated in two item descriptions. In embodiments, the system employs a neural network that includes a token encoding layer, an attribute summarizing layer, and a comparison layer. The token encoding layer accepts an item description as a token sequence and encodes the tokens with token attributes that are relevant to IPQ extraction. The attribute summarizing layer uses a convolutional neural network to generate a set of fixed-size feature vectors for each encoded token sequence. All feature vectors for both item descriptions are then provided to the comparison layer to generate the IPQ comparison result. Advantageously, the disclosed neural network model can be trained to make accurate predictions about the IPQ relationship of the two item descriptions using a small set of token-level attributes as input signals.