Resent-From: cbmvax!uunet!PICA.ARMY.MIL!protin
Resent-Message-Id: <9107032226.AA06164@relay1.UU.NET>
From: cbmvax!uunet!math.ucla.edu!jimc
Message-Id: <9006061844.AA23378@julia.math.ucla.edu>
To: lojban-list@snark
Cc: Eric Tiedemann <est@cs.nyu.edu>
Subject: Re: Nibbles on irresistible bait... 
In-Reply-To: Your message of "Tue, 05 Jun 90 04:35:51 EDT."
             <9006050835.AA02983@cs.NYU.EDU> 
Date: Wed, 06 Jun 90 11:44:24 -0700
Resent-Date:  Wed, 3 Jul 91 17:58:13 EDT
Resent-To: John Cowan <cowan@snark.thyrsus.com>
Status: RO

> Date:  Tue, 5 Jun 90 04:35:51 -0400
> To:  lojban-list@snark.uu.net, jimc, gls@Think.COM
> From:  est@cs.NYU.EDU (Eric Tiedemann)
> Subject:  Nibbles on irresistible bait...
   
> ... 
> la karen. djica le nu voha limna
> 
> "Karen wants the event of her swimming."
> 
> is structurally similar.  It's this that I'll use.  Now if I say,
> "mi go'i", and go'i has the obvious call-by-macro semantics, then
> the (genderless!) reflexive pronoun "voha" will refer to me.  I
> think it's possible to construct a consistent semantics in which
> a relative pronoun retains it's original referent even when that
> referent is replaced in a go'i repetition..but it would be
> sadistic!

This issue came up in -gua!spi.  There are a LOT of places where you
want an anaphor for a neighboring part of the sentence, and this is an
example of one of the most common: "X1 wants/ceases/quick/etc. to  do
(events) X2+1", the +1 signifying repetition (automatic in -gua!spi) of
the containing phrase X1 into the abstraction.  The -gua!spi doctrine
of anaphora is macro-like, i.e. you copy the words of the antecedent. 
(I'm pretty sure English produces a reference, i.e. the referent of the
antecedent, not the words, is referenced again; hence substitute
arguments are impossible.  I believe that lojbab said that Lojban uses
the macro form.)  Originally, once I bound an anaphor it stayed bound. 
But clearly 99% of the time the meaning desired is "I want for me to
swim" rather than "I want for Karen to swim".  Hence you have to unbind
SOME anaphora when copying, then rebind. This took a fair amount of
programming, particularly when it had to be idiotproof against
self-referential anaphora.  

Also, when there's a speaker change, does "mi" in the original mean the
original speaker?  The rule I came up with was that when you copy a
phrase, anaphora bound within the phrase are unbound while those bound
outside remain unchanged.  (Speaker assignments end up bound to an
outside copy even if set in a discursive on the same sentence.)  Thus: 
"mi djica lenu voha limna"; (new speaker) "go'i" means that the new
speaker is saying something like "So you want to swim", or "Sure, jump
in the lake" depending on discursives not shown here.  Whereas "mi
go'i" would produce the words "mi djica..." but their referent would be
the new speaker swimming.  It took quite a bit of work to decide
exactly what I wanted to do, and what matched the demands of the actual
texts (supposedly representative) that I wanted to translate.  

Now for a real challenge, try go'i to replicate some reported dialog. 
For example:

Says Kira:  And then I said to Karen, "I wish you would go jump".
Says Tami:  mi go'i

-gua!spi can reach into a main phrase with a modal predicate (i.e. "said")
to find that "you" is Karen, not Tami.  Thus Tami's sentence would mean
in English: "I said the same thing to her".  You can imagine the conflicts
between what the program was grabbing, what I thought was supposed to be
happening, and what the sentences "really" meant on closer examination.

		-- jimc