On Fri, 17 May 2019, 15:44 Jörg Rhiemeier, <[log in to unmask]> wrote:

> Hallo conlangers!
> On 17/05/19 02:34, And Rosta wrote:
> > On Thu, 16 May 2019, 22:49 Jörg Rhiemeier, <[log in to unmask]>
> wrote:
> >
> >> Hallo conlangers!
> >>
> >> [...]
> >> I think that successfully creating a conlang in which the Voynich
> >> Manuscript constitutes a meaningful text is pretty much equivalent to
> >> deciphering it, isn't it? After all, it is extremely unlikely that a
> >> text of that length makes sense in two utterly different languages!
> >>
> > Is that really so, when one of the languages is invented specifically to
> > fit the text?
> I think it would be an extremely unlikely coincidence if the text made
> sense in two completely different languages, so unlikely that it can be
> rejected out of hand! It is this very unlikeliness which makes
> deciphering extinct written languages possible - otherwise we couldn't
> know whether the language found in which the text makes sense is the
> right one or not!

"Coincidence" hardly seems the right word when one of the languages is
invented specifically to fit the text.

As you acknowledged in your reply to Gary, it is trivially easy to invent a
language that will generate a preexisting text of a couple of dozen words.
But you maintain that if the text were 30000 words (i.e. the VM) the task
of invention would be beyond human ingenuity, or at least that the
invention would entail inventing a language that in all relevant respects
is identical to the author's -- that generates the same sentences as occur
in the text.

> > (So if you arrive at a
> >> conlang in which the VM constitutes a meaningful text, you probably have
> >> guessed both the language and the meaning mostly right.
> >>
> > Can that really be so? Imagine you were faced with, say, a hundred pages
> of
> > illustrated Wolof text. Could someone sufficiently ingenious come up
> with a
> > grammar and lexicon that generates that orthographic text and has a
> meaning
> > appropriate to the illustrations? I'd guess yes. Is it likely that they
> > would have thereby enabled themself to read Wolof, to have understood the
> > Wolof text? Surely not. But you must think it would be nigh on impossible
> > to come up with a grammar and lexicon that generates that text.
> I don't know how you arrive at such a conclusion. Surely, if the text is
> long enough, the language found that way will be at least a more or less
> precise model of real Wolof, and will allow its builder to comprehend
> Wolof texts at least to some degree.

I look at how trivially easy it would be for a paragraph's worth of text
and then guesstimate that to extend that to 30000 words would be arduous
and require ingenuity but would not be insuperable; I imagine it to be
something one skilled and lateral-thinking conlanger could do in some
hundreds or thousands of hours. But that'd merely be my best guess. Is it
merely your best guess that the difficulty increases nonlinearly with text
length, or do you have reasons for thinking that?

Assuming that you hold that inventing a language to generate a preexisting
30000 word text would take essentially infinite hours (because you hold
that it could be done only by reinventing the authorial language, and I
hold that that is so utterly improbable as to be essentially impossible),
what text length would you guess is the shortest that one conlanger could
not invent a conlang for in a reasonable number of hours? I'm wondering if
we could then set it as a public challenge (a challenge for the public of

> Consider the process of deciphering ancient writings in an extinct
> language, such as Sumerian or Hittite. How does that work? It works by
> finding a grammar and a lexicon which make the old writings make sense.

Does it really? Surely more must be demanded than that, else decipherment
would be too easy. Need there not be additional reasons for supposing a
given decipherment to be correct other than that it merely works?

What is the smallest corpus that you think sober scholarship could deem
deciphered by production of a grammar and lexicon that generates it? Does
this mean that if a grammar and lexicon is produced that generates a given
corpus of that size or greater, and according to the lexicon the text
speaks of spaceships and photon torpedoes, the original authors of the text
were writing about spaceships and photon torpedoes?

Once such a linguistic system is found, the thing is considered
> deciphered and the found linguistic system is considered the language in
> which the texts are written (or at least a model of it). Of course, some
> room for error remains, but this gets quickly smaller and smaller the
> more text is successfully deciphered.
> > It is indeed more likely that it is written in an unknown language -
> either
> >> a lost natlang or a conlang. And
> >> it is perhaps not even encrypted, as the choice of the language alone
> >> makes it hard enough to crack!
> >>
> > But in that case, what sort of writing system is it? It does not look
> like
> > a writing system of any known sort (i.e. alphabet, syllabary, etc.), for
> > what possible values could the letters (given their distributional
> > characteristics) have that would yield something like a representation
> of a
> > linguistic utterance? Do you not agree that it is in fact very hard to
> > identify any linguistic characteristics of Voynichese, whereas any
> > nonencrypted text written in nonlogographic script does reveal many clues
> > about its linguistic characteristics, such as its morphological
> structure.
> The script is quite obviously alphabetic (it consists of 36 characters,
> 11 of which seem to be modifications of other ones), and it seems to be
> possible to distinguish vowels and consonants, even if the sound values
> aren't known (the EVA transcription convention[1] constitutes an
> educated guess). The glyphs look similar enough to the letters of the
> Latin alphabet to guess that the alphabet is related to the latter in
> some way.

If it is obviously an alphabet then it ought to be possible to make a
plausible proposal about what the values of the letters are. But (AIUI), it
isn't. There isn't (AIUI from a little casual light reading on the matter)
any readily discernible set of values that is phonotactically plausible let
alone consistent with the other distributional peculiarities of the letters
-- at least not if it's an alphabet and at least not if the language is
natural or at all naturalistic or pronounceable. Calling EVA an educated
guess seems a bit much; if something is an educated guess, shouldn't it be
conceivable that the educated guess might happen to be correct?