Re: anaphor means what? (was: oops! correction)

[cbmvax!uunet!math.ucla.edu!jimc]

> Date:  Mon, 22 Apr 91 12:17:31 EDT
> To:  lojban-list@snark.thyrsus.com
> From:  cbmvax!snark.thyrsus.com!cowan@uunet.UU.NET (John Cowan)
> Subject:  Re: oops! correction
   
> Lojban's position on anaphora is that pro-sumti and pro-bridi generally
> refer to the same referent as the selected sumti or bridi...
> 
> 	dei jitfa
> 	This sentence is false.
> 	.i go'i
> 	The previous sentence is false.
> 	.i go'e ra'o
> 	This sentence is false too.
> 
> The first sentence uses the utterance pro-sumti "dei" to refer to itself,
> and asserts of itself that it is false.  (Presumably, the guaspi version
> of this would macroexpand to an infinitely long string.)  

Gulp (red face...)  Fortunately my parser checks for being out of
memory. This self-referential business came up before, when I was
replicating a selbri into a subordinate clause that modifies it, and it
was cut off by a special hack; I didn't think imaginatively enough to
realize that legitimate user-supplied self-referential anaphora would
have to be handled effectively.

In the -gua!spi interpretation, by the time sentences 2 and 3 get hold
of sentence 1, sentence 1's anaphora have already been replaced by
their antecedents, and so (neglecting minor infinite loops) the same
effect is achieved as in Lojban.  This relies on the fact that repeated
sumti (without anaphora) will have the same referents.  

Heading off an objection: -gua!spi selbri are assumed to have implicit
anaphora for non-mentioned modal cases such as tense, which are
substituted when the selbri is digested, and the result of which is
copied when the subsequent anaphor is expanded. So the copied sumti /
selbri is interpreted in its original context.

The advantage of copying words and not referents is this:  When you get
to the phase of semantic analysis you are going to compute selbri
referent sets.  It is going to be a whole lot easier if the only thing
presented to the semantic analyser is a bunch of selbri.  If
additionally there are anaphora for about ten different grammatical
categories (categories that ought to be transformed away by this
point), it makes the semantic analyser much more complicated.  

Also, when you define an anaphor by copying words, you define it in
terms of grammatical structures being defined at or before that point.
Whereas if you tell a student "this will represent a copy of the
referents ultimately to be computed for this other structure", you are
reaching a long way down the process.  The student will have just as
much trouble understanding your teaching as the semantic analyser
program will.

Finally: the original motivation of this policy was to support modified
anaphora.  You hang a modal phrase, or even a "specified description"
(sumti on a sumti), on an anaphor and you expect it to replace or
supplement the original -- obviously the original words, since it's
outrageous to expect referents to be backmapped into symbols that
represent them, the modifications to be tacked on, and the symbols
reprocessed back into (certainly different) referents.  This insight
then led me to recognize the general advantage of copying words, not
referents.

James F. Carter        (213) 825-2897
UCLA-Mathnet;  6221 MSA; 405 Hilgard Ave.; Los Angeles, CA, USA  90024-1555
Internet: jimc@math.ucla.edu            BITNET: jimc%math.ucla.edu@INTERBIT
UUCP:...!{ucsd,ames,ncar,gatech,purdue,rutgers,decvax,uunet}!math.ucla.edu!jimc