From "decryption" to "translation"

[Anton]

At a glance, f68r1 has "or daiin" star, and f68r2 has "odaiin" star. But that's interesting... although purely morphological guesses will leave us with too much uncertainty. E.g., one could think of "in" as a tense selector, say, to turn a verb into the past tense, the "in" is appended. That would automatically mean that all vords ending with "in" are verbs. That's just an illustration, "odaiin" most certainly precludes that. ("odaiin" is the second most frequent Voynich star, not some outsider).

I'm thinking of some broader tests. One that I can think of is collecting a pool of texts and calculating the ratio of "inflections vocabulary size" to "vocabulary size". The former shall count all word forms, the latter - just unique words, flexion-invariant. This ratio would be greater than unity. Given a collection of such ratios for texts of supposedly comparable time and topics (herbals, medical books, astrology), we could then calculate the supposed vocabulary of the VMS in the assumption that the text is inflected. And see if the resulting figure looks reasonable.

Calculating the "inflexions vocabulary" of known texts and comparing that with the 8114 figure of Voynich would be the other way around, only we need to deal with normalization, because the overall volume of different texts differs.