Re: Stroke encoding?

[Jacques Guy <jguy@xxxxxxxxxxxxxxxx>]

27/02/02 09:41:26, Jorge Stolfi <stolfi@xxxxxxxxxxxxx> wrote:


>One major problem with that idea is motivation. There were no
>keyboards in the 1500s, so what would be the advantage of mapping a
>chinese character to a random-looking sequence of letters? 

Why, secrecy of course! And a very effective cipher for 
Chinese. Works irrespective of the dialect, you don't
have a long table of codes to look up, only a dozen or 
two of special codes, and with, just a little practice,
you can write it as fluently as the characters.

>That would
>perhaps make the script seem less intimidating to western students,
>but it would not help them much: they would still have to memorize the
>spelling of several thousand words, which would be completely
>unrelated to their sounds.

If they are the notes of a European (say, Venitian) missionary
who had learnt Chinese, and who wanted to keep them safe from
prying eyes, he had the perfect cipher, as uncrackable as DES.
And to those who knew Chinese, it would the perfect simple-
substitution cipher, actually steganographic: who would imagine
that _this_ could represent Chinese characters?

> Not to mention the fact that the encoding
>of a character would probably require twice as many strokes, and use
>up three times as much paper area than the character itself.

Yes, good point.

>Moreover, knowing the strokes and their order is not enough: in order
>to recover the character mechanically and unambiguously, one would
>also need to know the relative positions

Seldom necessary, in my experience. Very seldom

>(The only difference
>between the characters for "tu3" "land" and "shi4" "soldier" is the
>relative length of the two horizontal strokes).

Again, this is a very rare occurrence. No worse than
the phonetic spelling of Englisht shorthand (night = knight), 
not even bringing in the omitted vowels.

>The Voynichese words
>don't seem to be long enough to carry that information explicitly.

Also my feeling. The longest Chinese character requires 
27 strokes (it's cai2 "at last, still") but its common
form is only three strokes long: heng2, shu4-gou3, pie3


>There is another big problem with this theory, though. I don't see how it
>would give rise to the observed word structure -- which, on the other
>hand, is very much like the phonetic structure of Chinese syllables.

The only way to decide is to try. However, some very common
stroke combinations may have been encoded as single letter,
e.g. two and three horizontal strokes, the character kou3 "mouth",
and, or, stroke repetition may have been encoded with a 
letter indicating x2, x3, or a diacritic (a swirl, etc.).
The problem is to find a database that gives the stroke order
of every Chinese character. There must be one.

I am persuaded that this is not the solution, but it is
an idea well worth pursuing.

-------------------------------------------------------
Example. Take yan2 ("speech") for instance. It goes:

dot, long horizontal, short horizontal, short horizontal,
mouth. Total: 7 strokes.

The dot is often replaced in printing by a short 
horizontal. The distinction short/long can safely
be overlooked. We end up with  horizontalx4, mouth
or: dot, horizontalx3, mouth. Two to three letters.