From lojbab@lojban.org Thu Feb 22 16:49:24 2001 Return-Path: X-Sender: lojbab@lojban.org X-Apparently-To: lojban@yahoogroups.com Received: (EGP: mail-7_0_4); 23 Feb 2001 00:49:10 -0000 Received: (qmail 31112 invoked from network); 23 Feb 2001 00:49:09 -0000 Received: from unknown (10.1.10.26) by l10.egroups.com with QMQP; 23 Feb 2001 00:49:09 -0000 Received: from unknown (HELO stmpy-2.cais.net) (205.252.14.72) by mta1 with SMTP; 23 Feb 2001 00:49:08 -0000 Received: from bob.lojban.org ([209.8.89.72]) by stmpy-2.cais.net (8.11.1/8.11.1) with ESMTP id f1N0n2l79237; Thu, 22 Feb 2001 19:49:02 -0500 (EST) Message-Id: <4.3.2.7.2.20010222185345.00b9a4e0@127.0.0.1> X-Sender: vir1036/pop.cais.com@127.0.0.1 X-Mailer: QUALCOMM Windows Eudora Version 4.3.2 Date: Thu, 22 Feb 2001 19:52:50 -0500 To: Nick Nicholas , lojban@yahoogroups.com Subject: Re: [lojban] nilbroda In-Reply-To: References: <4.3.2.7.2.20010221150126.00b6d280@127.0.0.1> <4.3.2.7.2.20010216215932.00af0130@127.0.0.1> Mime-Version: 1.0 Content-Type: text/plain; charset="us-ascii"; format=flowed From: "Bob LeChevalier (lojbab)" At 03:08 AM 02/22/2001 -0800, Nick Nicholas wrote: > >> >What about "variadic" selbri like "(nil)du"? How does one know whether > >> >the last sumti belongs to ni or du? > > >>Ouch. Very good point. You either > >>(a) give up and don't admit nildu'o as a lujvo (which is no good); > >>(b) make nildu'o an exception, and have the ni2 precede the arguments of > >>du'o --- which leads to untold confusion; > >>(c) leave it as it stands, which is disastrous. > > >>One would say at this point that usage and context would always sort this > >>out. To which I answer that if I want English, I know where to find it. If > >>a lujvo has an ambiguous place structure, it has no business being a lujvo. > >A lujvo never has an ambiguous place structure. There may be ambiguity in > >the conventions for choosing a place structure, but they are aonly > >conventions. > >OK, Lojbab. What's the place structure of nildu'o? What is a ve nildu'o? I dunno. What does it MEAN? Remember that I always favored pragmatics over rules, and have never been enamored of the dikyjvo conventions (which I don't claim to understand). The fact that du allows an indefinite number of places does not mean that a -du'o lujvo has to, if it is not useful. But per the rules in your paper, nildu'o would have an indefinite number of places with ni2 being the last one. >If ni2 follows the seltanru places (c), as the Book says occurs for every >other ni- lujvo, then a ve nildu'o is a standard of quantity, if you're >comparing two things, and a comparandum, if you're comparing three things: > >xy. ni .abu dunli by. kei .y'ybu => xy. nildu'o .abu by. .y'ybu >xy. ni .abu dunli by. cy. kei .y'ybu => xy. nildu'o .abu by. cy. .y'ybu > >This place structure is ambiguous. Therefore it cannot be accepted as the >place structure of a lujvo. Therefore the rule that ni2 follows the >seltanru cannot hold here. Therefore we can only prepose it (b): > >xy. ni .abu dunli by. cy. kei .y'ybu => xy. nildu'o .y'ybu .abu by. cy. > >But you see, making one exception to the rule that semantic salience >dictates lujvo ordering opens up the floodgates. If ni2 should move up >because the places of du are unpredictable, that isn't far from saying ni2 >should move up in general because the places of any seltanru are >idiosyncratic. Of course. The floodgates were never closed. From the book, which I now have in hand: "What is said in this chapter represents guidelines, presented as one possible standard, not necessarily complete, and not the only possible standard. There may well be lujvo which are built without regard for these guidelines ..." (section 1) "... when we try to design the place structure of a lujvo ... it's given by the specific need we wish to express, and it determines the place structure of the lujvo itself. Therefore it is generally not appropriate to simply devise lujvo and decide on place structures for them without considering one or more specific usages for the coinage. ..." (section 3) Without a specific usage in mind, I would fall back on the observation that while du allows more than two places, such usage is rare. So thus, depending on how conventional I am feeling, I would opt for ni1 du1 du2 ni2 or ni1 ni2 du1 du2, with the latter possibly allowing open-ended du/n places. lojbab >As for a lujvo never having an ambiguous place structure, what happens when >you combine two gismu with variadic place structures? e.g. bunda 'pound', >which has as many degrees of subunit from x4 on as the local culture >allows, and dunli. What is the place structure of budydu'o? Produce specific usages, and come up with an answer. I would be unlikely to retain the variable places of du. Take budybunda, however. I cannot imagine its meaning lacking any culture information, so I therefore cannot resolve on a place structure. This does not bother me. >What is a xe budydu'o? > There's no way to combine the trailing places of bunda and the >trailing places of dunli so you know when the one starts and the other >ends. You'd better pray that you can drop one of those groups of places; >and since variadic places seem limited to units of measurement and dunli, >the seltanru seems safe to drop. But that's a messy contingency. Really, I >fail to see why we should have these variadics in the first place --- we >make do with ce-sets for just about everything else. Pity I didn't notice >this a decade ago... It is not the case that 2+2 = {4, 3+1} because sets do not equal numbers. At best you would have to use set selection with lu'i and say that it is one out of the set. But of course that structure was not well established until close to the baseline. > >The reason this topic came up is that it > >appears that the convention you chose for nil- lujvo does not match what > >you actually did > >As it turns out, the convention was not chosen by me, but by whoever had >written the abstraction paper before I'd written the lujvo paper. Well, Cowan wrote the abstraction paper. I just looked in the 5/94 version which indeed discusses nilklama with the current book convention, but your paper quoted previously does not seem to indicate the abstraction paper as a source, but rather an unspecified JL discussion of nunbroda. (Maybe it was an earlier version of the abstraction paper??) > >>I knew there was a reason I wanted those ni2 and traji4 in second position. > >>Oh well. > >If there is a reason, then state it, and we then know that under conditions > >applicable to that reason, there is an exception to the conventions. > >Once more: predictability is good, The dictionary is more predictable than the algorithm, so if we put it in the dictionary, no one needs to predict. The issue here is NOT that the lujvo is ambiguous, but that the book makes an ambiguous prediction about nildu'o. > and since each seltanru has its own >number of places, it would be nice to move recurring tertanru places which >would normally trail at the end of the predicate, where this is a regular >pattern. In reality, I wouldn't bother about ni2; the variadics do make it >impossible for ni2 to trail, So in the special case of the variadics there is a reason for ni2 to not trail. Fine. >It's all moot with the language baselined, But the lujvo place structure conventions, while baselined, are explicitly not definitive or binding, merely "predictive" as you might call it. >and it's not like I actually >care about nildu'o :-) , but an exception clearly has to be made for >abstraction lujvo involving variadics. And I do need to say my piece about >regular lujvo place structures: the notion that Lojbanists will memorise >idiosyncratic place structures for each of a zillion lujvo is utterly >doomed to failure. Yet we "memorize" more or less the meanings of a zillion English words, each of which has an idiosyncratic meaning for its prepositions usually but not always following a pattern. And like English, Lojban word meanings will ultimately be determined by usage, not by convention. The conventions allow us a starting point in determining place structure which will usually work (if only because we are trying to follow the conventions in creating the dictionary). If there are exceptions, then we have a variation from the default pattern, and there won't be a zillion of those. And of course no one memorizes ALL the words in the English language either. We guess meaning from context or we use a dictionary. Sometimes we guess incorrectly because a word is not predictable. The language still seems usable %^) > There will be regular principles for forming lujvo place >structures, or else they will have no place structures in practice: they >will be used for nothing but their x1s. And that will also likely be the case for many lujvo. >And the more those principles are followed, the easier the language is to >handle. I agree. But faced with an ambiguity in the principles, we have to make a choice. So we choose. And the choice might get written up in a forward to the dictionary (along with any other alterations or additions to the conventions in the book, which explicitly stated them to be incomplete.) The lujvo lexicon remains the only unbaselined portion of the language. That is why it is important to process those umpteen thousand lujvo and get place structures for them: to determine how binding the book conventions will be in fact, and whether we need to add alternates or additions to the conventions, or recognize that some words simply are idiosyncratic. The dictionary is the place to put any such additions, changes, alternatives. One reason I have not been hot on confining the lujvo work to only the most used words is that I suspect the problem cases that will define the language will show up in less-used words. But of course pragmatically, if we can only get 500 words analyzed without controversy, they will probably be the most used words, because those are what most people want to have resolved. lojbab -- lojbab lojbab@lojban.org Bob LeChevalier, President, The Logical Language Group, Inc. 2904 Beau Lane, Fairfax VA 22031-1303 USA 703-385-0273 Artificial language Loglan/Lojban: http://www.lojban.org