Recognition Error Correction Model Using Process Labels
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing voice and character recognition error correction methods do not provide a clear process for correctors to manually correct recognition errors, lacking transparency on the necessary correction steps.
Innovation Solution
A recognition error correction device that acquires pair data associating recognition results with process labels indicating correction processes for each word, and generates a correction model through machine learning to provide a clear environment for error correction.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of information
If manual correction of recognition errors is performed without showing the correction process, then correction work can be done, but the corrector cannot understand what kind of process has to be performed
Solution Approach 1:
The system provides feedback to the corrector by displaying the correction process information obtained through machine learning. The correction information display unit shows the correction process (such as deletion, substitution, or insertion) that the machine learning model predicts should be performed, enabling the corrector to understand and verify the correction steps before executing them.
2Extent of automation
If machine learning is used to generate a correction model, then automatic correction capability is improved, but the system complexity increases
Solution Approach 1:
The patent introduces an intermediary machine learning model that acts as a bridge between the recognition result and the final correction. This model learns the correction process from training data and outputs correction process information (deletion, substitution, insertion) that guides the correction without requiring complex direct control systems.
Solution Approach 2:
The system performs preliminary action by pre-training the machine learning model with correction data before actual correction operations. The model learns the correction process in advance from training sets containing recognition results and their corresponding corrections, so that during operation, it can automatically predict the appropriate correction process without real-time complex analysis.
Data Source
AI summary
An object is to construct an environment in which a process for correcting a recognition error of a recognition result of voice recognition or character recognition is shown. A recognition error correction device 1 includes a pair data acquisition unit 21 that acquires pair data in which a sentence of the recognition result of voice recognition or character recognition and a label string composed of process labels that are labels indicating a process for correcting a recognition error for each word constituting the sentence are associated with each other and a correction model generation unit 22 that generates a correction model that is a learned model for correcting a recognition error of the recognition result by performing machine learning using the pair data acquired by the pair data acquisition unit 21.