Refining optical character recognition results with natural language processing techniques
Srinivas Kumar Palvadi, Krishna Prasad K · IET conference proceedings. · 2025
The process also termed optical character recognition (OCR) involves converting images whichis in the form of print or handwritten data, obtained with the help of soft copy images, files as well as photographs, into text that can be read by machines. This technology has been pivotal in making the documents softcopy which helps to analyse and access. However significant progress, OCR still encounters challenges, particularly with illegible text or when the OCR software lacks robustness, resulting in inaccuracies in the converted text. This study aims to address these challenges by employing post-processing techniques over OCR results, the basic thing of utilizing Transformer models like BERT with the help of 2 step approach. The goal is to correct errors which improve the overall the overall quality of the file.