Ancient inscriptions are vital primary sources for understanding linguistic evolution, socio-political structures, and cultural history. The Halegannada script, used in early Kannada inscriptions, presents significant challenges for computational analysis due to its complex orthography, weathered stone surfaces, and limited annotated datasets. Conventional epigraphic analysis relies heavily on manual interpretation, making it time-consuming and difficult to scale. This paper proposes an end-to-end AI-driven framework for the automated recognition and translation of Halegannada inscriptions. The methodology incorporates image preprocessing techniques for noise reduction and contrast enhancement, followed by a hybrid Optical Character Recognition (OCR) approach that combines Convolutional Neural Networks (CNNs) for character feature extraction and Recurrent Neural Networks (RNNs) for sequence modeling, supported by Tesseract as a baseline OCR engine. The recognized text is then mapped to Modern Kannada and translated into English using Natural Language Processing (NLP) techniques. Preliminary experiments conducted on digitized inscription datasets from Mysuru University show encouraging recognition accuracy and translation consistency. The proposed framework supports digital preservation efforts, improves accessibility to historical inscriptions, and contributes to the field of computational epigraphy. Future work includes expanding training datasets and integrating transformer-based language models.
R et al. (Fri,) studied this question.