VMs: Model matches/mismatches...

[Nick Pelling <incoming@xxxxxxxxxxxxxxxxx>]

Hi everyone,

Just a quick thought: the VMs' system's reuse of common glyphs in multiple 
different ways seems quite likely to frustrate many automated analyses 
(such as HMMs).

For example, just look at "o" (though you could draw up similar lists for 
"y", or "d", etc):-
	at start of words (typically preceding gallows)
	in "qo"
	in "cho" / "sho"
	in "ol" / "or"
	in "eo"
	etc

What I find tantalising is that these motifs are so stylistically 
repetitive - yet while characterising them as pairs (as I have tried to do) 
seems to get close, it only gives half the story.

It should be clear from the above that I consider that using only the 
preceding glyph as context will fail to give a satisfactory account for 
much of the observed behaviour. What my four-column model (much like John 
Grove's OKOKO model) is trying to achieve is to capture this changing 
context through Voynichese "words" as they develop.

Even so, despite the fact that "we all know" what a Voynichese word looks 
like, a vast number of Voynichese words don't appear to fit any (currently 
proposed) word-level paradigm satisfactorily. Picking a random short 
paragraph from the CopyFlo (P2 on f43r), I've marked all "words" with "x" 
which my current model rejects (with an attempted explanation for each):-

x	pshesy		(model doesn't contain "p" or "f")
	otey
x	kshdy		(no transition from "k" to "sh")
x	opchdy		(model doesn't contain "p" or "f")
	kedar		(no transition from "e" to "d")
	okedy
x	chdy		(model requires "e/ee/eo" between "ch" and "dy")
x	shocphey	("cph" is missing from Column 3)
x	dytydy		(no match at all)
x	pchdy		(no match at all)
	kedy
	dam -
x	ytchedy		(no transition from "t" to "ch")
	chedy
	cheody
x	shy		(mode requires "e/ee/eo" between "sh" and "y")
x	qoiiin		(no transition between "qo" and "iiin")
x	sheeeky	(model has no "eee" - should be there, though)
	chedy
	dain
x	shy		(as before)
x	ykolor		(bad fit for whole word)
	otaiin
x	old -		(text fragmenting near end of line?)
x	dshedy		(no transition between "d" and "sh")
	qotedy
	dor
	cheey
x	odain =		(no transition between "o" and "d")

13 positive matches from 30 words is either a pretty bad hit rate (only 
43%) or a pretty good model (it matched nearly half of the sample), 
depending on how you look at it.

All the same, as I suspect that...
(a) the first four words are some kind of Neal key (because of their 
"p...p" structure)
(b) similarly, "shocphey dytydy pchdy" on the same line looks like a Neal key
(c) the model should contain "eee" (to handle "sheeeky")
(d) I could probably match "shy" and "chdy" without compromising the model
...I'm not *too* disheartened. :-)

BTW, I'm convinced that Neal keys do contain a different class of text from 
the rest of the text - though quite how that could be proved is another 
question entirely...

Cheers, .....Nick Pelling.....


______________________________________________________________________
To unsubscribe, send mail to majordomo@xxxxxxxxxxx with a body saying:
unsubscribe vms-list