Re: VMs: Re: Note for Jeff

[Nick Pelling <incoming@xxxxxxxxxxxxxxxxx>]

Hi Larry,

At 16:22 16/12/2003 -0500, Larry Roux wrote:
> I could show a few examples of where ch is NOT a single glyph 
> (unfortunately I am at work and don't have the pages readily handy).  So, 
> remember, your OPINION is that ch is a single entity.  Everyone does not 
> have to agree

Yes, I've seen those (perilously few) examples - but ultimately it's a 
judgement call as to whether they are statistically sufficient to make ch 
and sh composite glyphs. If you had a text (of an unknown language) with 
10000 instances of "m" and 10 instances of "n", would that be enough to 
assert that "m" is actually "nn"? Sure, that's more than a hapax legomena 
(ie a single instance in a text), but would it be sensible to draw the "nn" 
conclusion?

Perhaps more importantly for Jeff's position, it would be extraordinarily 
hard to construct a particularly persuasive argument that EVA "sh" (or, 
more accurately, "Sh") has much to do with the freestanding EVA "s" - yet 
because he seems to take EVA as literally true (when it's only a handy 
approximation on our long road forward), this is what he's assuming.

What I'm saying is this: relying purely on EVA (rather than looking at the 
text itself) would seems to me to be a superficial approach, and the kinds 
of problems you're discussing that arise from the transliteration of "ch" / 
"sh" / "s" (never mind all the "sh" variants... or might you argue that 
they are actually three letters?) are merely manifestations of that 
superficiality.

Of course, you *could* construct an argument for "ch" being two letters - 
and that *might just* make sense in a language (as in "qu" - if "u" only 
ever appeared after two letters, that is). However, in a cipher or an 
artificial language, that would make little or no sense, as the "h" would 
have virtually no information content (my guess is roughly 0.001 bits of 
information).

The postmodern methodology is to say "here's another opinion which we 
should add to our list - but we should not judge between competing 
opinions, as that would require us to take an Absolute Position, Which Is A 
Very Bad Thing Indeed." OK, there are *some* merits to this kind of 
approach - but the huge demerits are that you end up drowning in opinions, 
and that you are unable to resolve any issues of uncertainty that arise. 
Sounds familiar?

Few thinkers yet seem ready to say that being so absolutist about the 
poverty of absolutism is errant nonsense, and that we instead need to keep 
tolerance (divergent acceptance) and judgement (convergent evaluation) in 
dynamic tension. For example, we could spend our lives here simply 
documenting all the clever, interesting (and usually conflicting) opinions 
on the VMs which people have had - but what a waste of life that would be.

If we are to move forward here, we must do better than simply celebrating 
our intellectual diversity and ingenuity - we should instead try to find 
absolutes against which we can evaluate different opinions and ideas (such 
as using BIC for evaluating "word"-model fitness). Which VMs hypotheses and 
models are statistically significant, and which are simply rubbish?

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


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