Learning Data Selection Using User Operation Analysis

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Current methods for selecting useful learning data for machine learning, particularly in image processing, face challenges in distinguishing between correctly and incorrectly oriented images without manual intervention, due to privacy concerns and the labor-intensive process of manually labeling data, leading to decreased accuracy in orientation determination processes.

Innovation Solution

An apparatus and method that acquire and analyze user operations on data files to determine their suitability as learning data, utilizing a machine learning model to evaluate user reliability and data commonality, thereby automating the selection of useful learning data without manual data viewing.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If manual labeling of data files is performed to distinguish correctly and incorrectly oriented images, then the accuracy of learning data selection is improved, but the labor intensity and time consumption increase significantly

Engineering Contradiction:

Improveaccuracy of learning data selectionVSAvoidtime consumption for manual labeling

Core Design Contradiction:

Measurement precisionVSLoss of time

Solution Approach 1:

The system automatically analyzes user operations and determines data file suitability without requiring manual intervention. The determination unit autonomously evaluates whether data files should be adopted as learning data based on user operations, eliminating the need for manual labeling while maintaining selection accuracy

Inventive Principle:

Principle #25Self-service

Solution Approach 2:

The patent replaces the mechanical manual labeling process with an automated computer-based system. The analysis unit tracks user operations (viewing, editing, deleting) and the determination unit uses these operational patterns to automatically classify data files, substituting human labor with computational analysis

Inventive Principle:

Principle #28Mechanics substitution (Replace mechanical system)

2Reliability

If manual viewing of data files is performed to ensure data quality, then the reliability of learning data is improved, but privacy concerns arise and manual labor increases

Engineering Contradiction:

Improvereliability of learning dataVSAvoidprivacy concerns

Core Design Contradiction:

ReliabilityVSObject-affected harmful factors

Solution Approach 1:

The system introduces user operations as an intermediary indicator to assess data quality without directly viewing the data content. By analyzing operational patterns (viewing, editing, deleting) rather than the actual data files, the system maintains reliability assessment while protecting privacy

Inventive Principle:

Principle #24Intermediary (Mediator)

Solution Approach 2:

The patent replaces manual data viewing with automated operational analysis. The analysis unit monitors and evaluates user interactions with data files, using these operational metrics to determine data suitability without requiring human reviewers to actually view potentially sensitive content

Inventive Principle:

Principle #28Mechanics substitution (Replace mechanical system)

3Productivity

If automated selection of learning data is implemented without manual intervention, then productivity is improved, but the precision of distinguishing correctly and incorrectly oriented images deteriorates

Engineering Contradiction:

Improveefficiency of learning data selectionVSAvoidprecision of orientation determination

Core Design Contradiction:

ProductivityVSMeasurement precision

Solution Approach 1:

The system uses user operations as feedback signals to guide the automated selection process. By continuously monitoring how users interact with data files (viewing, editing, deleting) and using these operational patterns to refine selection decisions, the system achieves both high productivity and maintained precision

Inventive Principle:

Principle #23Feedback

Solution Approach 2:

The patent replaces manual visual inspection with automated operational pattern recognition. The system analyzes sequences and types of user operations to infer data quality and orientation correctness, substituting human visual judgment with computational analysis of behavioral data

Inventive Principle:

Principle #28Mechanics substitution (Replace mechanical system)

Data Source

PatentUS20240193921A1Apparatus for selecting learning data, method of selecting learning data, and non-transitory recording medium

Publication Date: 2024.06.13 PFU LTD

AI summary

An apparatus, a method, and a program stored in a non-transitory recording medium each of which acquires a data file, specifies an operation performed by a user on the acquired data file, and determines whether to adopt the acquired data file as learning data for machine learning based on the specified operation.