Re: Translation

[cbmvax!uunet!MATH.UCLA.EDU!pucc.PRINCETON.EDU!jimc]

> Date:          Fri, 6 Sep 1991 09:51:26 +1000
> From:  nsn@MULLIAN.EE.MU.OZ.AU
> Subject:       Re: Translation

> Hope this pleases you, because I advocate such dikyjvo analysis for all
> {rinka} compounds; save a hell of a lot of time. Before jimc starts huzzahing
> though, I reiterate what I, John Cowan, and, when he understands what Jim is
> talking about, Lojbab, have been saying: an analysis of all possible lujvo
> by an unambiguous dikyjvo test is not possible in Lojban. For starters, there
> is rarely a way of telling between the two major categories of lujvo, the
> broda be brode type ("transitive", of which the above analysis, "event ab-
> straction", is a special case) and broda je/joi/poi brode ("parellel").

Huzzah!  Huzzah!  You're quite right that it's hard to distinguish
parallel dikyjvo from the transitive and event abstraction types, and
I found I needed special grammar in -gua!spi to make the distinction
solid, grammar whose precise analog is not feasible in Lojban.  It will
take some work to figure out a suitable substitute in Lojban --
something that I don't feel comfortable to just go out and prescribe.

Lacking the grammar, you just have to "know", for example, that the
combination of a directional property and a motion word is parallel
while rinka and similar category members are never parallel.  I'd rather
have this distinction made by grammar, but it's better to do it word
by word than to have to make a creative judgement on every single lujvo.

You are also correct that the dikyjvo mechanism will be called upon to
handle a great variety of combination meanings.  But it surprised even
me to find out how much would straighten itself out simply and
automatically when:

    a.  Place structures are uniform over word categories and are
	designed to be most productive in dikyjvo.
    b.  Cleft place interpretation rules are used to import main-level
	sumti to interior abstractions and even "thing-type" sumti.
	Occasionally even x2 may be imported this way.  Your posydji
	example illustrates this point; speakers want to use the lujvo
	as a gismu-like relation without having to fill buried places
	by explicit words.
    c.  The same rules are allowed to run backward (retro-replication)
	to export buried sumti to containing bridi.

(By the way, "word category" is a small grenade.  For this purpose I
categorize pragmatically by typical behavior in dikyjvo, not
semantically.)

		-- jimc