Novel approach for digitization of ancient stone inscriptions and development of machine learning models for character recognition and transliteration

Abstract

Writing has always been natural mode of collecting storing and transmitting information down the ages Inscriptions are the earliest form of writing and have defined human culture down through the centuries through material remains They hold the key to information on the lives of our ancestors and trace human survival in ancient societies Different forms of inscriptions have been found through sources such as potsherds copper plates rock beds palm leaf manuscripts and walls encasing monuments Of these stone inscription have drawn remarkable attention in recent years from researchers and the government chiefly because they are authentic sources of information on history culture social welfare religious tolerance civil administration and medicine These inscriptions found in a typically conventional form are to be deciphered and translated into a contemporary language form in order for them to be read and their contents understood newlineStone inscriptions have been found in several languages including kannada Malayalam Urdu Sanskrit and steps taken to decipher them digitally Compared to these languages it must be noted that Tamil has possibly the largest quantum of inscriptions holding a wealth of information No digital system for Tamil however has been proposed so far According to the guidelines laid down by the Department of Archeology and the Tamil Virtual Academy Government of Tamil Nadu India a standardized Tamil alphabet came into existence in the 11th century, during the reign of King Raja Raja Chola Plenty of inscriptions are available for the period in question Hence the proposed work concentrates on Tamil stone inscriptions of the 11th century newline newline