Single Model Label Correction via Object Similarity

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing techniques for annotating training data for machine learning models require multiple prediction models, making it challenging to accurately correct prediction results when only a single prediction model is available.

Innovation Solution

An information processing apparatus and method that acquire a set of objects, evaluate the similarity between objects, and determine a label for a prediction target object based on similar labels predicted by a single prediction model for similar objects.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If multiple prediction models are used to correct prediction results, then annotation accuracy is improved, but device complexity increases

Engineering Contradiction:
Improveannotation accuracyVSAvoidnumber of prediction models
Core Design Contradiction:
Measurement precisionVSDevice complexity

Solution Approach 1:

The patent introduces an intermediary correction process that uses similarity evaluation between annotation targets to transfer label information from similar targets. This mediator mechanism allows a single prediction model's results to be corrected by leveraging information from similar instances, eliminating the need for multiple prediction models while maintaining high annotation accuracy

Inventive Principle:
Principle #24Intermediary (Mediator)

Solution Approach 2:

The patent copies label information from similar annotation targets to correct prediction results. By evaluating similarity between annotation targets and copying proven label patterns from similar instances, the system achieves accurate correction using only a single prediction model, rather than requiring multiple independent models

Inventive Principle:
Principle #26Copying

2Measurement precision

If manual annotation is performed, then annotation accuracy is improved, but productivity decreases

Engineering Contradiction:
Improveannotation accuracyVSAvoidannotation efficiency
Core Design Contradiction:
Measurement precisionVSProductivity

Solution Approach 1:

The patent enables the annotation system to self-correct prediction results by automatically evaluating similarity between annotation targets and transferring label information from similar targets. This self-service correction mechanism eliminates the need for extensive manual annotation while maintaining high accuracy, thereby significantly improving annotation productivity

Inventive Principle:
Principle #25Self-service

Solution Approach 2:

The patent performs preliminary similarity evaluation and label transfer between annotation targets before final label determination. By pre-processing and leveraging information from similar targets in advance, the system reduces the need for time-consuming manual annotation, thus improving overall annotation efficiency

Inventive Principle:
Principle #10Preliminary action

Data Source

PatentUS20250124313A1Information processing apparatus, information processing method, and storage medium
Publication Date: 2025.04.17 NEC CORP
  • US20250124313A1 patent drawing
  • US20250124313A1 patent drawing
  • US20250124313A1 patent drawing

AI summary

To determine, with high accuracy, a label to be given to an object even in a case where only a single prediction model exists, an information processing apparatus (1) includes: an acquisition unit (11) that acquires a set of objects; an evaluation unit (12) that evaluates a degree of similarity between objects included in the set of objects and identifies one or a plurality of similar objects which are similar to a prediction target object; and a prediction unit (13) that determines a label to be given to the prediction target object with reference to a similar label(s), the similar label(s) being a label(s) which is/are given to each of the one or a plurality of similar objects and which has/have been predicted by a prediction model.