| Sumario: | The Indus script has been one of the most perennial archaeological-epigraphical problems because of the lack of bilingual sources, extreme lack of data, and widespread physical destruction of the artifacts. New computational directions to overcome these limitations are provided by new developments in artificial intelligence, machine learning, and computer vision. The paper is a systematic review of AI-based methods for identifying, analyzing, and interpreting epigraphic texts and graphemes in damaged Indus Valley seal articles. After a systematic review of the literature, publications between 2009 and 2025 were thoroughly reviewed and classified into thematic groups, such as statistical and computational linguistics, deep learning-based visual analysis, allograph identification, pattern modeling, crowdsourced dataset construction and multidisciplinary methodologies. The review assesses methodological designs, data features, preprocessing procedures, model structures and reported performance measures, whereas common pitfalls include artifact distortion, small annotated data, segmentation ambiguity and computational capacity. The study indicates significant advancements in grapheme and motif detection, with ongoing shortcomings in the domain of semantic decipherment and linguistic interpretation. With comparison insights, a multi-phase analytic model is put forward that incorporates preprocessing, sign reduction, visual recognition, pattern recognition, and interpretation modeling. The paper concludes that significant developments should involve more intensive incorporation of computational practices into archaeological and linguistics to foster further research in digital epigraphy and computational archaeology.
|
|---|