Writing a fake Voynich by rules: copies, habits and the link between words

[Mauro]

I wanted to thank you for your interest in the skriba code, @mauro and @nablator.

Nevertheless, I am still skeptic about the use of such a generator. The results are quite good, but it has been a work of finetunning and adapting the rules to the skriba. I know that having a Voynich generator that works is quite difficult, but I really don't know id it is useful at all.

Yes, on this point I criticized Torsten's copy&modify and it's a weak point of your generator too (and of all generators). My feeling is that the parameters are too many in both cases.

But I think you made a great work. Sorry for summarizing (and for any mistakes), but you basically integrated trigram frequencies (including across spaces) with a template which filtered out unallowed words, and with a kind of selector which allows you to vary the proportion of important glyphs in the text ('t' vs.'k' etc.). Then a copy & modify mechanism, on a rather small percentage of words, reusing the same template filter as before. Finally, two 'gradients', one vertical and one horizontal along the page, controlling the distribution of particular characters ('p', 'f', or 'ch' 'd' etc. at line start). Wow, about everything known about Voynich stats all together.

Don't forget to post links to the official sample files your Skribe produced (ideally trained by sections, but by Currier A/B should do good too). I'd like to run some basic tests on my own and I'd like to examine the vocabulary, in particular for hapax legomena, because it's not yet clear to me if your filter template allows for the variety of forms of the VMS. But this is the bane of Voynich: strong rules coupled with a lot of variability make for big headaches.

If you think about the rules... I really don't think that they were the instructions to write the VMS. They are sort of the result of some other instructios (in thecase it was artificially written, no natural language). I have tried to summarize the rules into instructions but they don't make that much sense.... In THE_SCRIBE.md you can see "The instructions, as the master of the workshop might have given them"... do they really make sense? For those that don't have access to github, here are they....

I think the same. One can take an escape hatch and say the rules came naturally, unconsciously in the mind of the author (and the human factor can accomodate for the great variability too, double catch), but this borders on the unfalsifiable (it would anyway predict just one author). If it was so, we may never be able to know.