OCR and Translation System for Ancient Greek Scripts

Due to script deterioration, intricate ligatures, uncommon characters, and a lack of language resources, digitising and translating ancient Greek scripts is extremely difficult. Although they present encouraging alternatives, contemporary OCR and neural translation methods are still not optimal for processing old documents. Ten recent works on OCR enhancements, multilingual character recognition, semantic-aware translation, hybrid human-AI translation models, and integrated OCR-to-translation pipelines are examined in this study. Results show that transformer translation models in conjunction with deep learning-based OCR produce the best accuracy; however, domain adaption and error propagation continue to be significant problems. Although much progress has been made, the survey indicates that in order to attain dependable end-to-end performance, specialised OCR models, richer tokenization, and context aware translation frameworks are needed for ancient Greek scripts.