For automated document analysis, OCR (Optical character recognition) is a basic building block. The robust automated document analysis system can have impact over a wider sphere of life. Many of the researchers have been working hard to build OCR systems in various languages with significant degree of accuracy, character recognition rate and minimum error rate. Deep learning is the start of art technique with efficient and accurate result as compared to other techniques. Every language, moreover every script have its own challenges e.g. scripts where characters are well separated are less challenging as compared to cursive scripts where characters are attached with one another. In this chapter, we would take a detailed account of the state of art deep learning techniques for Arabic like script, Latin script and symbolic script.