Ginevra Rubergskier
Ginevra Rubergskier (born 1985) is an Italian-Estonian computational linguist and cryptographer affiliated with the Baltic Institute for Cryptography and Secure Computing (BICSC) in Tallinn, Estonia. She holds a doctorate from the University of Bologna (2012) on statistical methods for the analysis of unknown symbol systems, and subsequently worked at the European Centre for Linguistic Documentation in Vienna before joining BICSC in 2017.
Her research focuses on distributional approaches to undeciphered texts, particularly the application of information-theoretic and machine-learning methods to corpora for which no external key exists. She has described her interest in the Kristiansen corpus as "the cleanest test case I have found for methods that should not depend on knowing what the symbols mean."
Work
Rubergskier's paper "A dozenal primer hidden in plain sight: decoding arithmetic from a corpus of tagged tokens" (Language Codes 6, February 2024: 820–824) applied three quantitative methods to the text now known as the Dozenal Primer Inscription: positional bias tests, pointwise mutual information of adjacent sign pairs, and morphological productivity analysis of right-edge bundles. Working without proposed phonetic values, the study identified a stereotyped medial spine consistent with an equals marker, a binary operator with high internal collocation, a simplex unit consistent with ONE, and a productive derivational suffix generating multiples of twelve. Converging complement constructions, including forms consistent with 11+1, 10+2, and 6+6 all targeting the same derived token, were taken to diagnose twelve as the arithmetic base.
The paper explicitly evaluated and rejected a vowel–consonant phonographic interpretation of the same structural regularities, arguing that the rigid clause template and position-fixed derivational morpheme are more parsimoniously explained by an arithmetic register than by segmental phonology. A brief appendix lists nine diagnostic tests that future material could use to confirm or challenge the reading.
The paper has been well received within computational epigraphy. Camille Voudrin has described it as "a model of what distributional agnosticism looks like when it actually works." Inrik Üksküla notes that the AND_PLUS element Rubergskier identified in the Dozenal Primer Inscription appears with the same collocational profile in the Zagi Tablets, which he regards as "the strongest cross-corpus constraint we currently have."
Critics have noted that the arithmetic interpretation, while elegant, cannot be confirmed without either a bilingual or an independent identification of the object's cultural context. Rubergskier has acknowledged this limitation, calling the paper "a structural reading, not a decipherment."