Ranking Determination System Using Dual-Model Validity Scoring

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing ranking determination systems fail to accurately determine the ranking of multiple classifications estimated by models, as they only re-rank search results without evaluating the validity of the estimation results.

Innovation Solution

A ranking determination system that includes a processor to acquire multiple classifications based on a first model, calculate scores for the validity of these classifications using a second model, and determine a ranking for each classification based on these scores, thereby enhancing the accuracy of the ranking process.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If a single model is used to estimate classifications, then the processing is simple, but the accuracy of determining the ranking of multiple classifications is insufficient

Engineering Contradiction:
Improveaccuracy of ranking determinationVSAvoidmodel structure complexity
Core Design Contradiction:
Measurement precisionVSDevice complexity

Solution Approach 1:

The patent divides the classification system into two separate models: a first model that estimates multiple classifications for input data, and a second model that determines the ranking of these classifications. This segmentation allows each model to specialize in its specific function, improving the overall accuracy of ranking determination while keeping individual model complexities manageable.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent introduces a second model as an intermediary component that receives the multiple classifications from the first model and determines their ranking. This intermediary structure enables the system to evaluate and rank multiple classifications accurately without requiring the first model to directly perform ranking, thus resolving the contradiction between accuracy and complexity.

Inventive Principle:
Principle #24Intermediary (Mediator)

2Reliability

If only search results are re-ranked, then the process is simple, but the validity of estimation results cannot be evaluated

Engineering Contradiction:
Improvevalidity evaluation of estimation resultsVSAvoidsystem structure complexity
Core Design Contradiction:
ReliabilityVSDevice complexity

Solution Approach 1:

The patent implements a feedback mechanism where the second model evaluates the validity of classification estimates produced by the first model. The second model receives the classifications and their associated scores from the first model, processes this information, and outputs a determined ranking that reflects the validity evaluation. This feedback loop ensures reliable validity assessment while maintaining reasonable system complexity.

Inventive Principle:
Principle #23Feedback

Solution Approach 2:

The patent replaces simple re-ranking mechanics with a more sophisticated validation mechanism. Instead of merely re-ordering search results based on basic criteria, the system uses the second model to evaluate the validity of estimation results through learned relationships from training data, substituting mechanical re-ranking with intelligent validation.

Inventive Principle:
Principle #28Mechanics substitution (Replace mechanical system)

Data Source

PatentUS12056135B2Ranking determination system, ranking determination method, and information storage medium
Publication Date: 2024.08.06 RAKUTEN GROUP INC
  • US12056135B2 patent drawing
  • US12056135B2 patent drawing
  • US12056135B2 patent drawing

AI summary

Provided is a ranking determination system including at least one processor configured to: acquire a plurality of second classifications relating to second data based on a first model which has learned a relationship between first data and a first classification relating to the first data; acquire, for each of the second classifications, a second score relating to a validity of a combination of the second data and the second classification based on a second model which has learned a validity of a combination of third data and a third classification relating to the third data; and determine a second ranking for each of the plurality of second classifications based on the second score of each second classification.