From cbmvax!uunet!CUVMA.BITNET!LOJBAN Fri Oct 4 18:33:33 1991 Return-Path: Date: Fri Oct 4 18:33:33 1991 Message-Id: <9110042144.AA12338@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: VSO order X-To: lojban@cuvmb.columbia.edu To: John Cowan , Eric Raymond , Eric Tiedemann In-Reply-To: Your message of "Fri, 04 Oct 91 15:03:13 EDT." <9110041917.AA12524@julia.math.ucla.edu> Status: RO Lojbab writes, in response to Mark Shoulson responding to Dean Gahlon: > I'm going to interpolate what you are proposing as being that > > citka le nanmu le cripu > should be x1 x2 > and not the current x2 x3 with x1 unspecified > > in other words, you want an >unmarked< VSO order... Lojbab gives two reasons why Lojban needs the special pre-selbri x1 treatment. I thought it might be useful to outline how these problems are dealt with in -gua!spi, allowing the desired unmarked VSO. First, in an S-bridi (the bridi in a sumti), why should the first sub- sumti go in x2 rather than x1, which in Lojban is considered to be empty? Because it's not empty! A placeholder anaphor is there; the -gua!spi parser actually creates the word and sticks it in. The reason it's there is to let users know what the sumti means, like this (in -gua!spi, remember): Let's take "lo se citka be loi mlatu" representing "cat food", perhaps not optimally but it makes a good example. "citka" represents a relation between an eater and an eatee; such a relation is a list of pairs, the first member of each of which eats, and the second being eaten. Now retain only members consistent with subphrases, if any; in this case x1 has to be the mass cat (whatever that means, but that's another thread). One of the places will be occupied by the placeholder anaphor, either because the parser inserted it or the user said it explicitly (very rare). Collect the relation members in that place into a set, and modify according to the determiner / article "lo" (i.e. keep them all and unroll in extension). The result is the referent set of the sumti. The placeholder anaphor has that set as its referent, and in semantic analysis whenever you need the sumti's referents you link to the anaphor. Not to any other word in the sumti. Thus the placeholder anaphor has an essential role in semantic analysis. But being there, it makes natural the idea that any sub-sumti cannot occupy x1 (after conversion) because the placeholder is already in it. Lojbab's second justification for treating x1 specially is that very frequently in an abstraction there will be an "implicit" x1 which a common-sense listener will understand is the same as one of the main bridi arguments. For example, mi fengu le mlatu le nu citka lemi sanmi I'm angry at the cat for eating my food (meal) Clearly the eater is the cat, but in Lojban this sumti is invisible to the parser and so under an unmarked VSO policy, "lemi sanmi" would wrongly end up in citka x1. In -gua!spi the parser does know about the internal sumti because the dictionary entry for fengu (angry) tells it to replicate x2 into the x3 abstraction. Thus again, "lemi sanmi" goes into x2 because x1 isn't empty! This kind of replication instruction is very common, and even applies to x2 of the abstraction in a few cases, mainly comparisons. The speaker can, but rarely does, suppress replication by putting an explicit zo'e or other sumti in the target place. It is my opinion that the kind of common-sense implicit replication contemplated in Lojban is not robust enough to allow solid logical analysis of the resulting bridi; clearly specified algorithmic rules are necessary. This is why -gua!spi never is bothered by the place-assignment objections Lojbab makes to the unmarked VSO idea. Lojbab also points out, correctly, that the observative is a very basic kind of sentence and deserves low overhead support, perhaps more so than the unmarked VSO with explicit x1, which is what I am trying to make to work in Lojban. Nick made a suggestion, which I agree with, that the only problem comes up when an explicit x2 is present; under an unmarked VSO policy it needs "fe" just as under present policy x1 needs "fa", like this: Unmarked VSO Present policy English carvi carvi (no change) It's raining carvi fe mi carvi mi I'm getting rained on carvi le se baktu carvi fa le se baktu It's raining buckets So in the observative arena the burden is shifted from one not-so-common usage pattern to another similarly rare, while the burden is taken off of a pattern I use a lot in non-observative situations. -- jimc