From cbmvax!uunet!CUVMA.BITNET!LOJBAN Thu Sep 19 14:01:49 1991 Return-Path: Date: Thu Sep 19 14:01:49 1991 Message-Id: <9109191632.AA23075@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: dikyjvo To: John Cowan , Eric Raymond , Eric Tiedemann In-Reply-To: Your message of "Thu, 19 Sep 91 14:56:14 +1000." <9109190459.AA21414@julia.math.ucla.edu> Status: RO > Date: Thu, 19 Sep 1991 14:56:14 +1000 > From: nsn@MULLIAN.EE.MU.OZ.AU > Subject: dikyjvo > How about this as a discussion base for dikyjvo on lojban: > 1. the places of a lujvo should be a subset of the places of the component > gismu. > 3. places from a single given gismu in the lujvo, wherever scattered within > the lujvo, should preferably appear in the order they appear in the original > gismu. Yes, I fully agree with both. > 2. the places of the last gismu in the lujvo take precedence (appear first, > etc.) Yes with a caveat. In all the dikyjvo I use, x1 of the main term (the last gismu) appears first in the combination. x2 is occupied by an abstraction or a transitive sumti created from the pre-term, or is parallel to x2 of the pre-term (with very very rare special case exceptions as in batkla = bartu klama = exit, where bartu x2 is parallel to klama x3). Thus for x1 and x2 your rule is satisfied. (In double actor transitive relations like bapli "x1 compels x2 to do x3...", main term x1 and x2 are first, and the abstraction is in x3, again satisfying your rule.) The troublesome part is x3..x5 of the main term. In a transitive relation it is most common for the "real" action to be in the pre-term, like this: mi barzu'e le spati (banro zukte) I grow the plant (grow actor) But the definition of zukte is "x1 acts/does x2 purpose x3" so if the main term places took precedence you would have to skip over the purpose: mi barzu'e FI le spati I feel this is unnatural -- it's "obviously" a 2-place relation between the farmer and the plant, with other places like purpose trailing along afterward. Also people are likely to make errors counting main term extra places. On the other hand, if all the pre-term places come first, people have the same trouble accessing main term extra places. In -gua!spi I initially put all pre-term places first, but later decided this was wrong. I changed so main term places after the one involved in compounding were avoided like the plague, but if they were present they were important and hence should come before pre-term places. Thus I agree with NSN's rule but I warn that Lojban has far too many extra places on gismu common as main terms. Those places should be supported by BAI phrases. As Lojbab and I both point out, the design of dikyjvo echoes back in a major way to the semantics of gismu. > 4. the analysis of relation between gismu components in lujvo may be used > to restrict the meanings possible (there is nothing wrong with verboten-ing > the equivalent of adjective-noun from lujvo space - redhair != red hair. > Esperanto did this long ago). The analysis broda tai lo brode, however, has > been missing from existing formulations. This is certainly my position -- though on the specific example, note that I would interpret xunkre - xunre kerfa as a parallel dikyjvo whose referent is both red and hair. I believe that lujvo are so important and can be so useful that we should design their interpretation rules into the language. I also believe that the small number of gismu is a very important feature of Lojban, but if the users have to memorize the meaning of each lujvo as if it were a gismu, that advantage is lost. Thus my emphasis on algorithmically deriving the meaning of a lujvo from the component gismu. Note, however, that Lojbab wants to preserve semantic freedom to the maximum degree, as a component of the experiment testing the Whorf-Sapir hypothesis. If this freedom is indeed one of the design requirements of Lojban, then my algorithmic approach to lujvo is not compatible and will have to be abandoned. -- jimc