MODI Lipi Translator

Abstract

This project aims to develop a system capable of translating and transliterating MODI Lipi characters into Marathi Devanagari script. MODI script was the official script for Marathi until 1950. Although Devanagari has since become the official script for the Marathi language, MODI script holds significant historical value, as large volumes of MODI manuscripts are preserved in libraries and archival institutions across India. However, transliterating MODI into Devanagari is a challenging task because MODI documents are complex in nature and there is no standardized dataset available for experimentation. Character recognition in MODI script is particularly difficult due to its cursive, continuous, unconstrained writing style and the presence of numerous visually similar characters. The main objective of this project is to investigate the feasibility of an Optical Character Recognition (OCR)-based approach for converting MODI Lipi into Devanagari script. The proposed system evaluates OCR frameworks such as Calamari OCR, OCRopus, and Tesseract, along with deep learning approaches including VGG16 and Convolutional Neural Networks (CNN). The MODI-to-Devanagari conversion problem can be approached using either Natural Language Processing (NLP)-based or pattern-based techniques. This project focuses on a pattern-based approach for recognizing MODI characters and generating their corresponding Devanagari representations. The proposed system aims to contribute to the digitization, preservation, accessibility, and computational processing of historical MODI manuscripts.