Minoan and the Machines: Computational Approaches to the Decipherment of Linear A

In this paper, we review some of the significant literature surrounding the computational decipherment of Linear A, an undeciphered writing system from Bronze Age Crete (Aegean Sea), ‘hiding’ the so-called (unknown) Minoan language. Specifically, we summarize the methodology and results of two machine learning-based techniques that might be applicable to the study of Linear A, the first relying on modeling language decipherment as a minimum cost-flow problem and the second on comparing phonetic values across languages by using a generative framework. Additionally, we evaluate the effectiveness and limitations of these techniques and compare their efficacy and relevance to the decipherment of Linear A. We also review a consonantal cluster-based approach (a cryptanalytic ‘brute force attack’) and a feature-based similarity method, both of which run parallel to these machine learning techniques, and briefly discuss the advantages and shortcomings of the latter in comparison to the former. The aim of this paper is to provide our readers with a synthetic, but precise outline of new computational approaches for the decipherment of Linear A and to encourage further studies and research on this topic.