AI-EPIGRAPHY: An Interactive Tool for Computational Decipherment of the Indus Valley Script

The Indus Valley script, which dates back to the Bronze Age (2600–1900 BCE), is one of the most important ancient writing systems that has not yet been figured out. The tool employs corpus-based methodologies to discern potential language patterns, utilizing frequency distribution analysis, locational examination, and transitional probabilities among signs. Some of the most important features are statistical significance testing (Z-test) to look for differences from expected sign distributions, collocation analysis to look at contextual relationships, and n-gram modeling (1-4 grams) to observe patterns in sign sequences. A multinomial naive Bayes classifier, which is a type of machine learning, predicts semantic interpretations based on established inscriptions and proposed readings. We have showcased both the input methods with the help of case studies and the HCI insights. The tool looks at claims about proposed linguistic structures, like prefixing and suffixing systems, and claims about script characteristics, like logographic vs. syllabic qualities. Codebase is available at https://github.com/atulsharma0071/indiahci2025.