A Framework for Pre Processing, Recognizing and Distributed Proofreading of Assamese Printed Text

Abstract

The paper provides a brief outline of the framework established for digitizing, recognizing and proofreading printed Indic documents with Assamese as a target language. The establishment of such a framework is essential as it depicts the workflow for the digitization and archival of the scanned text and it has a high impact on the end result. The main idea behind the framework is to build the foundation for an automated text correction engine which provides suggestions based on the experience set generated using manual text correction procedure and machine learning approaches. Most of the works already done in this domain is based on the dictionary approach which has its own shortcomings like inability to correct real-word errors, redundant queries, large size, non-exhaustive collection etc. Hence, in this research, the dataset will be built from the scratch based on the experience gathered during digitization which in-turn shall contribute in increasing the accuracy of the OCR engine by means of post-processing.