A Comprehensive Analysis of Approaches for the Digitization of Text using Optical Character Recognition

With the ever-advancing field of data analytics and text mining, it is becoming quite evident how structured textual data can help enhance the amount of useful information that can be extracted from text material of all kinds. Unfortunately, a large amount of text data is only available in the offline form or stored in the form of images of text which makes it harder to be utilized for any kind of machine learning-based analytics. One of the ways to make this data productive is by converting it to online text by the means of Optical Character Recognition (OCR.) A tremendous amount of research has been done in the past about the multitude of ways OCR can be applied for recognizing structured text from unstructured handwritten or digital text. In this paper, we explore this domain of text recognition and deep learning called OCR and provide an understanding of how this landscape has changed over the years while also answering some vital questions regarding the kinds of algorithms, datasets, and motivations behind these OCR research models