ML-Based Document Data Correction System

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Conventional OCR engines fail to accurately extract data from remittances due to poor image quality, leading to errors and the need for manual correction, which is time-consuming and inefficient.

Innovation Solution

A Machine Learning (ML)-based computing system that receives documents, scans for mis-captured data fields, uses historical correction data to determine deltas, and applies these patterns to correct the data fields, automatically replacing incorrect data with accurate information.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Extent of automation

If conventional OCR engines are used to extract data from remittances, then automated information extraction is achieved, but data accuracy deteriorates due to poor image quality and fundamental limitations in the OCR engine

Engineering Contradiction:
Improveautomated information extractionVSAvoiddata accuracy
Core Design Contradiction:
Extent of automationVSMeasurement precision

Solution Approach 1:

The system uses historical correction data as feedback to train an ML model that predicts and corrects OCR errors. The model learns from past correction patterns and applies them to correct current extraction errors, continuously improving accuracy through feedback loops.

Inventive Principle:
Principle #23Feedback

Solution Approach 2:

The patent replaces the mechanical/conventional OCR extraction system with an ML-based correction system. Instead of relying solely on traditional OCR accuracy, the system uses machine learning models trained on historical correction data to predict and fix extraction errors automatically.

Inventive Principle:
Principle #28Mechanics substitution (Replace mechanical system)

2Measurement precision

If manual correction of errors is performed by clients, then data accuracy is improved, but productivity deteriorates due to time-consuming correction processes

Engineering Contradiction:
Improvedata accuracyVSAvoidcorrection efficiency
Core Design Contradiction:
Measurement precisionVSProductivity

Solution Approach 1:

The system enables self-service correction by automatically detecting and correcting extraction errors using ML models trained on historical correction data. The system serves itself by predicting and fixing errors without requiring manual intervention, transforming a manual process into an automated self-correcting system.

Inventive Principle:
Principle #25Self-service

Solution Approach 2:

The system performs preliminary correction actions by pre-training ML models on historical correction data before actual error correction is needed. The model is prepared in advance with correction patterns from past errors, enabling rapid automatic correction when new errors occur without requiring manual analysis each time.

Inventive Principle:
Principle #10Preliminary action

3Measurement precision

If manual correction of errors is performed by clients, then data accuracy is improved, but loss of time increases due to tedious correction tasks

Engineering Contradiction:
Improvedata accuracyVSAvoidcorrection time
Core Design Contradiction:
Measurement precisionVSLoss of time

Solution Approach 1:

The patent replaces the manual mechanical correction process with an automated ML-based system. The machine learning model automatically predicts and corrects errors based on trained patterns, substituting human manual correction with automated intelligent correction that is both accurate and time-efficient.

Inventive Principle:
Principle #28Mechanics substitution (Replace mechanical system)

Solution Approach 2:

The system changes the parameter of correction methodology from manual human intervention to automated ML-based prediction. By transforming the correction process into a computational parameter change problem, the system achieves both high accuracy and reduced time loss through automated pattern recognition and application.

Inventive Principle:
Principle #35Parameter changes

Data Source

PatentEP4325447A1Machine learning based system and method for facilitating correction of data in documents
Publication Date: 2024.02.21 HIGHRADIUS CORP
  • EP4325447A1 patent drawingFigure 1
  • EP4325447A1 patent drawingFigure 2
  • EP4325447A1 patent drawingFigure 3

AI summary

A system and method for or facilitating correction of data in documents is disclosed. The method includes receiving one or more documents from and scanning the received one or more documents by using a document processing system for obtaining one or more mis-captured data fields. The method further includes obtaining a historical correction data and determining one or more deltas based on the obtained historical correction data by using a trained data correction-based ML model. Further, the method includes parsing the determined one or more deltas into one or more datasets, generating one or more correct data fields corresponding to the one or more mis-captured data fields and automatically replacing the one or more mis-captured data fields with the generated one or more correct data fields based on one or more predefined rules.