Hi everyone,
At 17:51 27/11/2004 +0100, J.Siemons wrote:
Came around this, from the University of Tilburg ,The Netherlands,
Oct 20,
Something with the VMS and insect language.....
http://arxiv.org/PS_cache/cs/pdf/0406/0406054.pdf
Wonder if I do understand a lot of it, Oh well.
AIUI, Dr Paijmans' paper asks whether we can transcribe/notate bee
dances so as to look for Zipf-Law-like behaviour, & hence to see
whether it has language-like behaviour. Though the author doesn't
actually follow this idea through, he/she feels sufficiently confident
by the end to conclude that it doesn't.
Despite citing Gabriel Landini, I think that Paijmans hasn't really
learnt the overall lesson of the VMs' encounter with Zipf's Law, which
is that its presence/absence is a weak correlative factor in areas of
uncertain "languageness" (like the VMs), & not really solid enough
ground to build a proof upon.
Also, Paijmans clearly flags an information-centric bias (just like
Gordon Rugg's), which is based on (what I would call) an innately
positivistic view of communication, where a transmitted signal must be
*certain* (ie perfectly quantized and perfectly precise, AKA "the map
*is* the territory"). However, in the context of bee-dances, it is
nonsensical to say that a single dance codes to...
5 cybernetic units (sic!) as to direction, 4 to 5 as to
distance and 2 to 3 as to the number of workers needed. This
totals to about 12
bits, equivalent to a human language of 4000 phrases
(signifiants with corresponding
signifi´es), needing less than a hundred words by human or
english standards.
Put differently, a code of all possible combinations of only
three characters would
cover the communication system of the honey bee dance. [p.3]
This misses the key difference between real-world languages and
computer science grammars: the former operates under conditions of
uncertainty, the latter under conditions of certainty. Redundancy is
built into the heart of human languages in order to overcome the
mishearings & misinterpretations of real-life interpersonal
communication, much like error-correcting codes: computer grammars
(and, I guess, universal languages) operate in a different situation
entirely. One might just as validly ask, if (as Shannon demonstrated)
the sequential letter-to-letter predictability (ie, the negentropy) of
English texts is so informationally low, why do we bother to
transcribe using 26 letters?
In fact, the key issue skirted by Paijmans' paper is how one should
best transcribe bee choreography given that we don't actually
understand how its mechanisms works - trying to recast the problem in
terms of "information content" (in much the same way that Gordon Rugg
does) is actually quite unhelpful.
The analogy I'm trying to draw with the VMs should now be fairly clear
- the "bee-dance" of Voynichese is something we all "understand", but
transcribing it should ultimately only be a means to understanding the
underlying mechanisms behind the behaviour.
Cheers, .....Nick Pelling.....
______________________________________________________________________
To unsubscribe, send mail to majordomo@xxxxxxxxxxx with a body saying:
unsubscribe vms-list