RE: VMs: No stats no fun

[Nick Pelling <incoming@xxxxxxxxxxxxxxxxx>]

Hi John,

At 08:38 09/08/2003 -0400, John Grove wrote:
>         You found recently that lots of words 'could' start with the pair 
> 'dy' if
> you chopped up the transcription at the appearance of 'e' or 'ee'. This
> really is not new news - it's just a different way of saying that lots of
> words in the VMS token-A style end with 'dy'.No matter what transcription
> set you use - you will still observe patterns like in B-token pages there
> are lots of lines beginning with an 'l' character/glyph while in the A-token
> pages this doesn't happen.

> You could choose to ignore the line as a separate
> unit for your stats, but that doesn't make this anomaly go away.

It's a fair cop - there are indeed (as you noted) a handful of patterns and 
statistics which remain relevant to the kind of transcription I'm pursuing. 
But it is only a handful, out of a large sack of results.

Also: note that once you tokenise out <ol> and <al> pairs from the 
balneological section, you find freestanding <l>s appearing within the 
lines (such as in <qol>s). But generalisations about language differences 
need hard data: in the [Currier A, I believe] page f45r, for example, there 
are three line-start <l>s, but no free-standing <l>s to be seen (FWIW I 
think that the <l> on the last line is part of a split-up <ol> pair).

> If the
> author used some sort of pairing as you suggest - why does the alignment on
> line starts in B-token pages differ so dramatically from those on A-token
> pages.

Can you please explain what you mean by "alignment"? (Thanks!)

> If we assume that there is no difference between A & B, and that the
> statistical distinctions that have been noted to date is purely coincidence
> then we still have to develop a reason for the patterning to show itself.

I'm not claiming that the statistics generated to date are meaningless or 
irrelevant - any theory (whether analytical or generative) would need to 
explain (or forensically reproduce) them, just as much as any other 
statistic. What I *am* claiming is that the statistical ambiguities 
observed may have arisen from the process of looking for primary properties 
amongst (largely) secondary artefacts - where any faint signal would likely 
be swamped by noise.

That's not so much "coincidence" as the result of (I suspect) conscious 
misdirection on the original author's part. I believe we need to give him 
more credit... :-)

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


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