Deciphering the Indus Script: A Computational Linguistic Reconstruction of a Proto-Dravidian Abugida through Comparative Analysis of the Sangam Old Tamil Corpus

This study presents a computational linguistic investigation of the Indus script under a Proto-Dravidian hypothesis, using the Sangam Old Tamil corpus as a later comparative linguistic proxy. The method analyzes recurring and non-recurring Indus sign sequences and compares their statistical distributional structure with Old Tamil two- and three-character n-grams. Candidate correspondences are subsequently evaluated using positional-context analysis, frequency controls, robustness testing, and linguistic interpretation.

The study proposes a phonetic and compositional model in which some Indus signs may encode syllabic or multi-phoneme units and in which graphical modification may correspond to phonological modification. Particular attention is given to recurring sign families and to proposed readings including Mahadevan signs #47, #48, #53, #99, #328 and #342. The urn-sign series is provisionally interpreted with #328 as a base aa value, #99 as ai, and the graphically expanded #342 as the composite aiya hypothesis.

The resulting sign values are applied to selected Indus seal inscriptions and the Dholavira signboard. Because the Indus inscriptions substantially predate the Old Tamil corpus, reconstructed expressions are treated as possible earlier Dravidian forms whose closest attested reflexes may occur in Old Tamil; the study does not claim that Mature Harappan inscriptions were written in Sangam-period Tamil.

This Zenodo record includes the research paper together with computational source code, Indus inscription data, Old Tamil corpus inputs, derived n-gram datasets, and supporting files intended to facilitate reproduction and independent evaluation of the computational results.