From cbmvax!uunet!CUVMA.BITNET!LOJBAN Sat Nov 2 12:53:58 1991 Return-Path: Date: Sat Nov 2 12:53:58 1991 Message-Id: <9111011735.AA05135@relay1.UU.NET> Reply-To: cbmvax!uunet!MATH.UCLA.EDU!pucc.PRINCETON.EDU!jimc Sender: Lojban list From: cbmvax!uunet!MATH.UCLA.EDU!pucc.PRINCETON.EDU!jimc Subject: Re: Lojban duplications X-To: lojban@cuvmb.columbia.edu To: John Cowan , Eric Raymond , Eric Tiedemann In-Reply-To: Your message of "Thu, 31 Oct 91 19:40:00 EST." <9111010047.AA02872@julia.math.ucla.edu> Status: RO Art Protin writes: > While I will accept that some of the add-on-places are really > restrictive clauses, I am not receptive to a limted, bounded, definitive > list other than the numbered places. My understanding of lojban/Loglan > NEVER had this hangup about unfilled places. A recent post of mine shows what a pain it is to be stuck with a rigidly specified set of "essential" places -- yet I feel it isn't possible to claim Lojban is a "logical" language unless at least some effort is made to specify precisely what the predicate relations are. Or to state the contrapositive: Lojbab says that in order to be a "real" language, a conlang has to allow that the places of gismu, and both places and core definition of lujvo, are as renegotiable as a government contract, so that (jimc's inference) it's impossible to state in a public or global manner what the predicate relations are. Therefore it's impossible for a "real" language, constructed or not, to be a predicate language. I'm not saying that the conclusion is wrong (yet); but I don't like it one little bit. Nonetheless, it's clear to me that there are too many numbered places on some gismu (e.g. the transport means on some but not all motion words) yet essential places leap out at you that you never would have thought of when writing the dictionary (e.g. river of WHAT, or the transport means on cadzu, where I really needed it to make "crawl"), and intermixed with these are the pervasive context and supplementary places which are clearly not essential parts of the relation (e.g. speaker ID, or language of expression, or exemplar, or I think tenses). Bob Chassell's point is well taken that the predicate relation is the key to predicate languages. Art's is well taken that you need to be able to decline to state the occupants of various places. And Lojbab's point is well taken that if we're going to use the language in real life, we can't afford to get bogged down with infinitely many junk places. When dealing with problems like this one, I find it very useful to make a mechanistic computational model -- the database or Prolog model is very applicable to predicate languages, and has the advantage that eventual AI code can be modelled 1-1 on the theory. When you phrase the theoretical arguments in terms of the model, you can then test consequences and have clear results which you can evaluate. In particular, the debate about unfilled places can be phrased in terms of, what variable, anonymous or otherwise, belongs in the vacant places, and what do you do with referent set members that match the prototype thus constructed? A question like that is much easier to discuss, than to discuss the "reality", such as it is, of Lojban or natural languages. -- jimc