folk song lyric texts & entropy statistics

[RenegadeHealer]

I agree with you guys. I think using a proposed crib or key to turn a specimen of the proposed plaintext language into the ciphertext is a good early step for seeing if you’re at all on the right track. It’s a crude test with decent negative predictive value: if no specimens of the target language (written by a fluent speaker contemporaneous to the ciphertext, naturally) look anything like the uncracked ciphertext, that’s a pretty reliable sign you’ve got at least the wrong key, if not the wrong plaintext language, but probably both.

Of course this is a small sample of text compared to the VMs, so it’s hard to extrapolate meaningfully. But in my estimation your model passes this first of many hurdles, Geoffrey, and is therefore worth exploring further. I don’t have JKP’s eye for valid vs. invalid Voynichese, though, so Icould be wrong about this. Might you be willing to display the output of your reverse translation into one of the available Voynichese fonts? I think that might make it easier to see how Voynichy it feels to read it.

For what it’s worth, arguments still rage over just how passably Voynichy Torsten Timm’s bot’s output is; it’s not a set of arrangement rules whose higher level properties are easy to fake.