N-gram Review Analysis for Language-Independent Recommendation Accuracy

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Existing product recommendation systems face limitations in accuracy due to dependence on user knowledge, language and grammar dependencies, and inability to account for foreign language reviews and misspellings, leading to exclusion of valuable predictive information.

Innovation Solution

A computer-implemented system and method that uses n-grams from user reviews to predict target user preferences, correlating reviewer identities with n-grams and ratings to provide language-independent recommendations for unrated products, leveraging the full predictive value of reviews without requiring specialized programming for spellings or translations.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If word-based predictive systems are used to recommend products, then language-dependent predictions can be made, but foreign language reviews and misspelled terms are excluded losing valuable information

Engineering Contradiction:

Improveprediction accuracyVSAvoidforeign language and misspelled review data

Core Design Contradiction:

Measurement precisionVSLoss of information

Solution Approach 1:

The patent introduces n-grams as an intermediary representation that bridges the gap between raw review text and prediction models. By converting reviews into n-gram sequences and using cosine similarity to compare n-gram patterns, the system can match reviews across different languages and spellings without requiring explicit translation or spelling correction, thus preserving valuable information while maintaining prediction accuracy

Inventive Principle:

Principle #24Intermediary (Mediator)

Solution Approach 2:

The system transforms the prediction approach by changing from word-based matching to n-gram-based pattern matching. This parameter change allows the system to capture semantic similarities in reviews regardless of language or spelling variations, as n-grams capture contextual patterns rather than exact word matches

Inventive Principle:

Principle #35Parameter changes

2Adaptability or versatility

If peer-to-peer recommendation systems are used, then personalized recommendations are provided, but recommendation accuracy depends on recommender knowledge and user incentive is difficult

Engineering Contradiction:

Improvepersonalization capabilityVSAvoidrecommendation accuracy

Core Design Contradiction:

Adaptability or versatilityVSMeasurement precision

Solution Approach 1:

The system enables self-service recommendations by automatically analyzing review patterns and generating predictions without requiring active user participation in the recommendation process. Users simply need to provide reviews, and the system autonomously processes this data to generate personalized recommendations based on n-gram pattern matching

Inventive Principle:

Principle #25Self-service

Solution Approach 2:

The patent introduces an intermediary computational layer that objectively analyzes review data using n-gram patterns and cosine similarity, removing the subjective element of human recommenders while maintaining personalization. This intermediary system processes review data systematically to generate accurate predictions without relying on recommender knowledge or user incentives

Inventive Principle:

Principle #24Intermediary (Mediator)

3Productivity

If ratings-correlation systems are used, then recommendations are generated by comparing user ratings, but the trust value is limited because independent users rating highly may have different reasons

Engineering Contradiction:

Improverecommendation generation efficiencyVSAvoidtrust value

Core Design Contradiction:

ProductivityVSReliability

Solution Approach 1:

The system introduces n-gram pattern analysis as an intermediary that goes beyond simple rating comparison. By analyzing the actual text content of reviews through n-gram patterns, the system can determine whether high ratings from different users are based on similar reasons, thereby establishing trust in the recommendations

Inventive Principle:

Principle #24Intermediary (Mediator)

Solution Approach 2:

The patent replaces the mechanical rating-comparison system with a more sophisticated text-based analysis system. Instead of merely comparing numerical ratings, the system uses n-gram extraction and cosine similarity to analyze the semantic content of reviews, providing a more reliable basis for trust in recommendations

Inventive Principle:

Principle #28Mechanics substitution (Replace mechanical system)

Data Source

PatentUS11748797B2System and method for providing recommendations to a target user based upon review and ratings data

Publication Date: 2023.09.05 UNIVERSITY OF TULSA

AI summary

A method and system for recommending online content based upon the predictive value of reviews provided by third party reviewers correlated to the identity of those reviewers. The correlated data is used to predict how a target user would rate an unrated item, and to make recommendations to the target user.