OCR-PostProcessing
Loading...
Date
2024
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
The OCR post-processing project involves the improvement of text generated by text recognition.
The goal is to achieve the best possible improvements using various approaches. For
implementation four components are required:
The backend, which tries to enhance the given texts using three different approaches, including
two dictionary methods and one AI-based improvement. Additionally, the backend calculates
statistics to assess the performance of each system.
Improvements are tested and visualized through the frontend, which allows for the uploading
of multiple files or entire datasets at once to be improved with all desired correction systems.
All changes are displayed after the improvement process is completed and the backend-calculated
statistics are visualized.
An API connects the user interface and the backend. In Python, using Flask, several endpoints
were defined to facilitate the exchange of information between the frontend and the backend.
Furthermore, a test pipeline allows for the improvement to be used and tested without a
frontend. This pipeline can process a predefined folder structure, correcting all files contained
and comparing them with the ground truth.
The result is a comprehensive application composed of Python modules, JavaScript code, and
PowerShell scripts.