Re: CONLANG: Click

[John Cowan <cowan@LOCKE.CCIL.ORG>]

{Doug Merritt} {write*} :

> Hmm, you've got a ways to go. You haven't specified any means
> whatsoever of defining the sets that underlie Click. You made
> a vague appeal to set theory, but you can't strictly use set
> theory because...well, a bunch of stuff, for instance, there is
> no "universal set" (the catch-phrase is "there is no universe of
> discourse") in set theory.
> 
> You fiddled a bit with appealing to English translations without
> actually requiring English, but that doesn't work either...e.g. what
> does "in" mean? There's a bazillion meanings; practically one
> per verb phrase.

Consider the following text from the -gua!spi report:

# What Definitions Mean 
# 
# A  predicate word expresses a relation between the occupants  of
# its cases.  In English and all natural languages,  words are
# "defined"  by a sentence or two; the words in those sentences are
# often defined circularly in terms of the word being defined.  In -gua!spi,
# on the other hand,  the text definition is merely a learning aid.
# The  predicate is  actually defined by a set of all  thus-related
# object lists.  For example,  the referent set of "eats" includes a
# member with our example rat  in first case and our example cheese
# in second,  as well  as  numerous  other members containing other rats,
# foods, and so on ad (almost literally) infinitum.
# Other predicates  like {cu=pair} have referent sets
# that  are  actually infinite.
# 
# Language users are not expected to be familiar with every object set
# that was, is now or ever shall be thus related.  A big part of language
# behavior consists of the listener adding to his knowledge of which
# items are thus related, which information the speaker sends to him.
# Each person has his own limited experience of the world, but we speak
# of "the referent set" of a word independent of a person because words are
# supposed to mean the same thing to each person, that is, if a person is
# aware of a particular referent set member, typically he will agree with
# other language users which word's definition it is a member of.
# 
# Humans are very good at generalizing from a few referent set members so as to
# recognize novel referents, and they are not satisfied with a word until they
# can do such a general recognition algorithm and usually come out with the
# same answers their neighbors do.  But mechanical users of -gua!spi cannot be
# expected to show such skill, and neither can beginning human users such as
# infants.  They must build up a referent set for a word by exhaustively
# hearing referent set members.  If an advanced human, or advanced software,
# can transcend the official definition of -gua!spi words, that's fine -- a
# common (but risky) strategy for humans will be to use their native language
# as a guide to -gua!spi meanings.  However, -gua!spi words are still defined
# officially in terms of referent sets simply because this definition is known
# to be tractable both for theory and for practical implementation.  A -gua!spi
# referent set is perfectly suited to be represented as a Prolog database, if
# truncated to a practical size.

Now admittedly there are known problems with this extension-only view of
predicates:  "x1 has by nature a heart" and "x1 has by nature kidneys"
come out to be the same predicate.  (I say "by nature" to eliminate
fiddling objections anent partly dissected x1's.)

Nevertheless, it isn't hopeless.  Set theory paradoxes don't really apply
to languages that talk only about the real world, because the real world
contains only a finite number of things, and therefore at most denumerably
many classes of things.  Jacques drags in the Frege definition of numbers,
which complicates matters somewhat, but not if he sticks to only finite
Frege integers.

-- 
John Cowan					cowan@ccil.org
		e'osai ko sarji la lojban.