Evidence · Etruscan: the readable language that cannot be fully translated
The best Etruscan-to-English model trained on 2,891 translated inscriptions reached BLEU 10.1.
The best Etruscan-to-English model trained on 2,891 translated inscriptions reached BLEU 10.1.
What it means
Key questionHow far can the combinatory method and computational analysis extend the roughly 250 securely understood Etruscan words?
2 answers, none ahead yet
Openunchanged
Answer this evidence bears onMachine-learning translation trained on the existing translated inscriptions cannot go beyond the formulae that are…
Supports1 for, none against
Still possibleunchanged
Next steps 1
1
Run a distributional (co-occurrence and position) analysis over the Larth/OpenEtruscan corpora restricted to pre-1991 ET texts, propose meanings for frequent unglossed words, and score them on the c. 1,100 texts added in ET 2014 as a held-out set.
~20 h
—not judged
Ranked by how likely each is to move the key question, for the time it takes
Sources 2
Machine Learning for Ancient Languages: A Survey
SRCSommerschield et al. 2023
Cite it
Mysteries Department. (2026). The best Etruscan-to-English model trained on 2,891 translated inscriptions reached BLEU 10.1. [Evidence EVD0DEECND3S6S0]. https://mysteriesdept.com/r/EVD0DEECND3S6S0
@misc{mysteriesdept_evd0deecnd3s6s0,
title = {The best Etruscan-to-English model trained on 2,891 translated inscriptions reached BLEU 10.1.},
author = {{Mysteries Department}},
year = {2026},
note = {Evidence, catalog key EVD0DEECND3S6S0},
url = {https://mysteriesdept.com/r/EVD0DEECND3S6S0}
}