Re: VMs: RE: f78r pairified...

[Nick Pelling <incoming@xxxxxxxxxxxxxxxxx>]

Hi John,

At 11:53 04/08/2003 -0400, John Grove wrote:
> Couldn't you also say that all the words end in 'd' and the same stats
> would result with the first letter of every word being 'y' (without
> pairifying).

It's another possibility: but note that I came to the idea of {e/ee/eee/eo} 
being an end-of-word marker from a different question: how to compactly 
(and secretly) encode a language's declensions & conjugations? 
[Essentially, I think this is what Graham W is reaching towards].

Even given a special token (I think <ch/sh>) for abbreviation, (most?) 
languages would still have the issue of how to decline/conjugate words 
without giving the game away. My insight was that you'd probably need a 
*combining mechanism* - ie one with a combinatorial aspect - to do this, 
and the only common-enough (and obvious) feature of the language that would 
seem to be able to achieve this is the {e|ee|eee|eo} regexp.

Certainly, in the transcription I made, the second most frequent word start 
"pair" was <d> (8/86, ie 9.3%), which is consistent with the idea that "d" 
(following {e|ee|eee|eo}) is a null. But then again, if <d> is a null, then 
the stats point to <y>'s also being a null - hence my suggestion that <dy> 
is a null (or rather, "a paired-cipher representation of a single null token").

> You could probably take that farther and say that sh/ch/t/k are actually the
> end of word markers and most words begin with 'e'.

Alternatively, if the letters in each word are written back-to-front, then 
sh/ch/t/k might be the start-of-word markers, and the 'e' family the 
end-of-word markers. There are still lots of permutations to frustrate us 
yet. :-o

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


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