[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: VMs: Work on the relation penstroke -> letters?
At 08:34 13/01/2004 +0100, Elmar Vogt wrogt:
Yet most people seem to take the current tanscription schemes for granted,
only give a fleeting glance to this question which I feel is very basic and
fundamental. So, did I miss research which clearly answered that question, or
are people simply taking the transcription for granted, since it's easier to
tackle with the statistical apparatus we have?
You're absolutely correct - EVA isn't (what GC would call) a
glyph-transcription, but is rather (closer to) a stroke-transcription.
Given the ambiguity over what-is-or-isn't-a-glyph, this is so that you can
construct your own glyph-transcription (for example, if you think iiin is a
glyph, pre-process it so that *is* a glyph) for doing your own statistical
For example, looking at "raw" EVA word lengths - though relatively
interesting - is fairly likely to be absolutely misleading, because of
distortion "ch", "sh", "cfh", etc (add your preferred glyph-set here).
And as for verbose cipher candidate pairs (like "qo", "ee", "dy", "ol",
"or", to name but five), these appear so frequently that, if the VMs *is* a
cipher, then many of our most cherished statistical insights (as to letter
adjacency, letter frequency, apparent vowels/consonants, word length, etc)
may be largely worthless.
Unfortunately, the "why-not-use-EVA" (ie "the-map-is-the-territory")
assumption is being actively used by a number of people (most notably Jeff,
but I suspect others as well), which (I'm afraid) probably only serves to
add noise to the overall signal here. :-(
Of course, any statistical analysis of the VMs' text should detail the
exact transcription and glyph-set used for pre-processing (for
reproducibility), as this is a key modelling assumption - but this
important step occasionally gets overlooked in the excitement. :-(
Cheers, .....Nick Pelling.....
To unsubscribe, send mail to majordomo@xxxxxxxxxxx with a body saying: