A Computational Decipherment of Linear B

Paxton-Fear, Katie ORCID: https://orcid.org/0000-0002-4472-8955 (2018) A Computational Decipherment of Linear B. In: Computer Applications and Quantitative Methods in Archaeology Conference (CAA 2018), 19 March 2018 - 23 March 2018, Tübingen, Germany.

Preview

Presentation Download (1MB) | Preview

Abstract

The mystery of the world’s undeciphered scripts continue to draw in amateurs and experts alike. Machine translation currently allows us to translate known languages and writing systems such as between English and French or between English and Greek but this technology could be used to unlock the mysteries of the past and computers may hold the key to the understanding of these scripts. This project is a reconstruction of the decipherment of the Linear B script using computational techniques. Computational decipherment is a very important field; many ancient languages remain undeciphered such as Linear A, Etruscan, Rongo Rongo in addition to other ciphers or unknown writing system such as the Voynich Manuscript. Advances in decipherment could reveal more about the ancient world and answer key archaeological questions. In addition to deciphering more languages decipherment techniques can be used to further enhance machine translation by allowing translation with smaller corpuses. This project has deciphered Linear B in 3 stages, like the original decipherment, with the assumption that we have all tablets available in the dataset. In the first stage, an unsupervised algorithm will be used to automatically find inflected nouns. In the second phase, the results of this algorithm will be used to create a grid of values where signs share vowels or consonants. In the final phase, we will use a genetic algorithm to suggest possible values using evidence from statistics and known values of similar and related signs.