Content Analysis Using Multiple ML Models for Higher Confidence

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing machine learning models struggle to maintain high levels of confidence in real-world content analysis, particularly for fine-grained tasks involving subtle visual distinctions, despite achieving excellent results on benchmark datasets.

Innovation Solution

A combinatorial decision-making approach using independently configured ML models to enhance overall confidence by combining their individual, relatively low-confidence results, employing conditional logic and hierarchical classification architecture.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Device complexity

If a single ML model is used for fine-grained content analysis, then the model complexity is low, but the confidence level in outputs deteriorates

Engineering Contradiction:
Improvemodel complexityVSAvoidconfidence level
Core Design Contradiction:
Device complexityVSReliability

Solution Approach 1:

The patent combines multiple independently configured ML models to perform content analysis. Each model independently analyzes the content item and provides a confidence score. The results are merged through a combinatorial decision-making process that aggregates confidence scores from multiple models, thereby achieving higher overall confidence levels while maintaining reasonable model complexity for each individual model.

Inventive Principle:
Principle #5Merging (Combining)

2Reliability

If multiple ML models are used for content analysis, then the overall confidence level improves, but the system complexity increases

Engineering Contradiction:
Improveoverall confidence levelVSAvoidsystem complexity
Core Design Contradiction:
ReliabilityVSDevice complexity

Solution Approach 1:

The system segments the content analysis task into multiple independent model evaluations. Each ML model operates independently as a separate module, analyzing the content item and producing individual confidence scores. This segmentation allows the system to achieve higher reliability through multiple independent assessments while managing complexity by keeping each model relatively simple and independent.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The system implements a feedback mechanism where confidence scores from multiple models are aggregated and combined to produce an overall confidence level. This combinatorial decision-making process provides feedback that enhances the reliability of the content analysis by leveraging agreement among multiple independent models, while the structured feedback aggregation keeps system complexity manageable.

Inventive Principle:
Principle #23Feedback

3Ease of manufacture

If existing ML models are used without modification, then the implementation cost is low, but the performance on fine-grained tasks deteriorates

Engineering Contradiction:
Improveimplementation costVSAvoidfine-grained analysis accuracy
Core Design Contradiction:
Ease of manufactureVSMeasurement precision

Solution Approach 1:

The patent combines the outputs of multiple existing ML models to achieve enhanced fine-grained analysis capability. By merging the confidence scores from several models that each perform the same fine-grained content analysis task, the system achieves higher measurement precision and reliability without needing to modify or retrain the individual models, thereby maintaining low implementation cost.

Inventive Principle:
Principle #5Merging (Combining)

Data Source

PatentEP3807783B1Content analysis
Publication Date: 2025.08.13 QUBEEO LTD
  • EP3807783B1 patent drawingFigure 1~2
  • EP3807783B1 patent drawingFigure 3
  • EP3807783B1 patent drawingFigure 4A

AI summary

In one aspect a computer-implemented method of processing a content item to extract information from the content item comprises steps of: receiving the content item at a processing stage; and determining whether the content item satisfies a predetermined content condition, by i) providing content of the content item to a plurality of content analysers, each of which applies machine learning content analysis thereto, in order to make an independent determination of whether that predetermined content condition is satisfied, and provides a confidence score for that independent determination, and ii) making an overall determination of whether the content item satisfies that predetermined content condition based on the confidence scores provided by the content analysers for their respective independent determinations.