[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: VMs: Grove words

On Wed, 29 Sep 2004, Jorge Stolfi wrote:
> Yes and no. My grammar actually generates all the VMS words, Grove
> words included. However, the topmost rules are
>   Word:
>     33827 0.96296 0.96296 NormalWord
>      1301 0.03704 1.00000 AbnormalWord
>   AbnormalWord:
>       716 0.55035 0.55035 Multiple
>       213 0.16372 0.71407 GroveWord
>       372 0.28593 1.00000 Weird
> The grammar proper (with the crust-mantle-core division and all that)
> hangs from category NormalWord, which accounts for 33827 tokens (96.3%
> of the text). The categories GroveWord, Weird, MultiCore,
> MultiCoreMantle, EmbeddedAIN, and EmbeddedYQ are merely long lists of
> words, most of them with occurrence count 1. Together, these account
> for 1301 tokens, or 3.7% of the total.
> The category GroveWord could be defined as G.NormalWord, where
> category G yields the four gallows.  However, that would introduce
> ambiguous parsings for words like EVA "kody" (is that a GroveWord
> "k"+"ody", or a NormalWord "kody"?)
> As a computer scientist, I have been pavlovized to avoid ambiguous
> grammars; and that prejudice is the explanation for several
> questionable choices in my paradigm. A linguist would probably have
> cared more about simplicity and less (or not at all) about unique
> parsings, and produced a shorter grammar.

Me, too, as a linguist and a computer scientist - I have to eat - though I
think on both accounts one is allowed to avoid ambiguity by means of a
rule system that prioritizes choices.  Presumably kody would end up a
Grove Word.

However, I can see why this ambiguity is unsatisfactory and unsatisfying -
because otherwise the Gallows class enters into the grammar as a Core
constituent, i.e., in the middle of the middle.

So I wonder if it's possible to juggle the grammar to fix this somehow,
rather than resolving the problem by fiat, as it were.  I can see that
this would involve a major bit of surgery, however, and, unfortunately, I
haven't got any particularly insightful suggestions at the moment!

Perhaps things would look different if one separated the i- and e-
elements from the flourishes?  One might divide the gallows as well, to
allow at some future date for a way to handle "split gallows."  To avoid
the issue of which flourishes go together without straying too far from
EVA one could recode, e.g., m, d, t, p as im, ed, t1, t2, for example.
I'm inclined also to suggest coding ch and sh as ehe and eshe, using s
for the s-flourish, and h for the bar-ligature.  Maybe chc and cshc would
be safer.

I have the impression that however one goes about making word-internal
syntactic sense of Voynichese it complicates matters if one is trying to
simultaneously explain "glyph-internal" syntax and the syntax of the basic
orthographic units.  That is, I think that interglyph spacing does not
match interletter spacing.  Some letters hang from or depend upon others,
creating a single "cursive" or "glyphic" mass that embodies several
letters.  A charming way to misdirect the attentions of non-initiates,
though perhaps its mere chance.

My suspicion is that if the glyphs are treated as the basic units one is
in the position of an alien student of English trying to say
simultaneously that h tends to follow c, s, t, etc., in ch, sh, th and
that aeiou tend to follow bcdfgh etc.  Two levels of detail are being
confounded.  Up to a point this is possible in a single set of rules, but
it would be more awkward if a and e, but not iou were written as a
ligature with a preceding h.

Of course, if the internal morphology of the words is
transparent one might be noticing that com comes before pound and fort,
and not than C comes before V.  Either the morphology of the underlying
language or the encryption scheme - if any - could hide that.

Of course, if EVA m, d, t, ch, sh, etc., really are units, then any
syntactic analysis will stubbornly treat recoded im, ed, t1, ehe, eshe as
units, too.  As it is, byt not associating i and e flourishes, they will
flock to i and e respectively.  But I hope that unleashing them will
clarify matters to some extent.

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