Aim:

          Optical character recognition (OCR) can be used for the online retrieval of the printed material such as medical documents, forms, or applications for retrieving valuable information that was available in the printed documents. Deep learning approaches have been used to solve natural language problems.

Abstract:

           According to a recent study, the COVID-19 pandemic continues to place a huge strain on the global health care sector.As a result, the amount of digitally stored patient data such as discharge letters, scan images, test results or free text entries by doctors has grown significantly.This medical data does not conform to a generic structure and is mostly in the form of unstructured digitally generated or scanned paper documents stored as part of a patient’s medical reports. This unstructured data is digitised using Optical Character Recognition (OCR) process. A key challenge here is that the accuracy of the OCR process varies due to the inability of current OCR engines to correctly transcribe scanned or handwritten documents in which text may be skewed, obscured or illegible.The proposed work uses a deep neural network based self-supervised pre-training technique,Robustly Optimized Bidirectional Encoder Representations from Transformers (RoBERTa) that can learn to predict hidden (masked) sections of text to fill in the gaps of non-transcribable parts of the documents being processed. Evaluating the proposed method on domain-specific datasets which include real medical documents, shows a significantly reduced word error rate demonstrating the effectiveness of the approach.

Proposed System:
          More recently, neural networks and deep learning approaches have been used to solve natural language problems. Post-correction methods have been particularly developed and applied such as auto-encoders on Twitter and Wikipedia corpus which had promising improvement in the accuracy with appropriate settings like word lengths and type of the lexicon used to find the nearest match to the incorrect word or neural text embeddings. Equally Long short term memory (LSTMs) have been used for character-aligned strings or the Bidirectional Long Short Term Memory Networks (biLSTMs) to produce a robust character-based language model which does not require annotated training data.

Advantage:

         Using the proposed method, the authors were able to reduce the error rate from 21.2% to 4.2% in the document. Our approach uses the pre-trained model which is not as computationally intense as the approach proposed by Bassil et al. given that our approach is not accessing a tremendous database.