Single Model Label Correction via Object Similarity
Find Innovative SolutionsGenerate 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
Engineering Contradiction Analysis
1Measurement precision
If multiple prediction models are used to correct prediction results, then annotation accuracy is improved, but device complexity increases
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
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
2Measurement precision
If manual annotation is performed, then annotation accuracy is improved, but productivity decreases
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
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
Data Source
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.