Re: TEI element for a grapheme?

Am 28.09.2016 um 01:24 schrieb Martin Holmes:
> Yes, this works nicely for a single language group using a single
> script. But there are so many cases which are not so simple. Consider this:
>
> :-)
>
> I typed three "characters" to create a combination sign. In my email
> client, and in my mind, they are distinct characters (as if I'd typed
> "Yay!"). In your client, the software might substitute U+1F600 GRINNING
> FACE, which is a single glyph, and which you might perceive as a single
> character; someone else might receive exactly what I sent, three glyphs,
> and yet still "think of" it as a "character" because they use emojis all
> the time.
>
> It just seems to me that any attempt at a definition which depends on
> the perception of a non-specific, presumably non-expert user is hardly a
> definition at all.

"Eppur si fa", i.e. despite all these possible complications, every text 
editor makes a decision about what is a "grapheme" (although not all of 
them may agree on a single definition).

A grapheme is what is skipped when the user pressed the back arrow 
button. (A more strict version uses the backspace button instead of the 
back arrow).

In my computer, configured with my locale, using the composition editor 
of my mail client, "è" is grapheme, even though I typed it as COMPOSE + 
"a" + "`".

If I write "sarà" (typing "s", "a", "r", COMPOSE, "a", "`") and the 
cursor is at the end of that word ("sarà|"), pressing the back arrow 
will move the cursor over one grapheme, leading to this situation: 
"sar|à". Pressing the back arrow will not move the cursor between "a" 
and "`": I will not find my self in this situation "sara|`", No Italian 
speakers would want that behaviour (and my locale is it_IT).

(To see the difference between the definition with the back arrow and 
that with the back space one has to resort to Hangul, but the discussion 
is already complex enough with the Latin alphabet.)

This example is not meant to show that the editor's definition is the 
definitive definition of the term "grapheme". But it is an 
_implementable_ definition, although locale/context dependent. It is 
even in Perl [1]. In the case of a TEI transcription that is going to be 
processed though computer programs, I think that one should limit the 
discussion to the definitions that can be described in 
machine-implementable steps. Definitions that take into account 
individual preferences are hardly machine-implementable.

Regards,

[1] 
http://www.perl.com/pub/2012/05/perlunicook-string-length-in-graphemes.html

-- 
Gioele Barabucci <[email protected]>