Grammar Rule Generation Using Review Clustering and Synthetic Data

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Existing topic extraction and aspect-based sentiment analysis models in natural language processing require extensive human-generated grammar rules, which are impractical for large datasets like electronic catalogs with millions of items, leading to insufficient identification of parts of speech and interdependencies.

Innovation Solution

A machine learning-based grammar rule generation system that clusters reviews by rating (positive, negative, neutral) and uses human annotators to create rules, allowing the system to train a grammar classifier, receive feedback, and improve classification efficiency.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If human-generated grammar rules are used for topic extraction and sentiment analysis, then measurement precision of text features is improved, but device complexity and loss of time increase due to extensive manual input requirements

Engineering Contradiction:

Improvetext feature extraction accuracyVSAvoidtime for human input and rule generation

Core Design Contradiction:

Measurement precisionVSLoss of time

Solution Approach 1:

The system performs preliminary clustering of reviews by sentiment (positive, negative, neutral) and automatically generates grammar rules from clustered data before actual topic extraction and sentiment analysis. This preliminary rule generation reduces the need for extensive manual rule creation during operation, thereby reducing time loss while maintaining precision.

Inventive Principle:

Principle #10Preliminary action

Solution Approach 2:

The system uses unsupervised learning to automatically cluster reviews and generate grammar rules without requiring extensive human intervention. The model serves itself by learning patterns from raw review data and creating its own grammar rules, significantly reducing the time and effort needed for manual rule generation while maintaining extraction accuracy.

Inventive Principle:

Principle #25Self-service

2Measurement precision

If human-generated grammar rules are used for large datasets like electronic catalogs, then measurement precision is improved, but productivity decreases due to impractical manual input requirements

Engineering Contradiction:

Improvetopic extraction accuracyVSAvoidprocessing efficiency for large datasets

Core Design Contradiction:

Measurement precisionVSProductivity

Solution Approach 1:

The system automatically clusters millions of reviews and generates grammar rules through unsupervised learning without requiring manual rule creation for each dataset. This self-service capability enables the system to handle large electronic catalog datasets efficiently, maintaining high topic extraction accuracy while achieving practical productivity for large-scale processing.

Inventive Principle:

Principle #25Self-service

Solution Approach 2:

The system changes the approach from manual rule generation to automated clustering-based rule generation. By transforming the input data into clustered groups and deriving grammar rules from these clusters, the system achieves both high precision in topic extraction and improved productivity for processing large datasets with millions of items.

Inventive Principle:

Principle #35Parameter changes

3Measurement precision

If extensive human-generated grammar rules are used, then identification precision of parts of speech and interdependencies is improved, but device complexity increases

Engineering Contradiction:

Improveparts of speech identification accuracyVSAvoidsystem complexity for rule management

Core Design Contradiction:

Measurement precisionVSDevice complexity

Solution Approach 1:

The system automatically generates grammar rules through unsupervised clustering of review data, eliminating the need for complex manual rule management systems. The model self-organizes the grammar rules based on learned patterns from clustered reviews, maintaining high precision in parts of speech identification while reducing device complexity by removing manual rule creation and management infrastructure.

Inventive Principle:

Principle #25Self-service

Solution Approach 2:

The system performs preliminary unsupervised clustering of reviews to identify patterns and generate grammar rules before actual text analysis. This preliminary organization of data into meaningful clusters simplifies the subsequent rule application process, maintaining high identification precision while reducing the complexity of rule management during operation.

Inventive Principle:

Principle #10Preliminary action

Data Source

PatentUS12572734B1Machine learning grammar rule generation

Publication Date: 2026.03.10 AMAZON TECH INC

AI summary

System and methods are provided that generate machine learned grammar rules. The system clusters a sample set of reviews. A human annotator manually generates grammar rules for the clustered sample set. Synthetic data is generated from the manually generated grammar rules and text. The system combines the sample set, the manually generated grammar rules, and the synthetic data. A grammar classifier is training using the training data. A grammar classifier receives input text and output a predicted grammar rule. The predicted grammar rule is used for natural language processing.