From LOJBAN%CUVMB.BITNET@uga.cc.uga.edu Tue Nov 29 20:48:01 1994 Message-Id: <199411300147.AA05925@nfs1.digex.net> Date: Tue Nov 29 20:48:01 1994 From: Gerald Koenig Subject: Re: diversity Status: RO JL: >la djer cusku di'e >Better use a real experimental cmavo, which are of the form xV'V. >I think the only ones currently in use are xe'e and xa'e, so you can >pick any of the other 23 possibilities. (All the ge'V are already >taken anyway, so it doesn't make much sense to use ge'x.) GK: Thanks for the explanation. I wasn't really happy with those kludges xe'eway. I'll say xa'a (the first vowel) for the begin scope marker and xu'u (the last vowel) for the end scope marker. >I suppose your cmavo would be in selmaho LAhE, that changes one sumti >into another type of sumti. GK> It is broader than just a sumti changer. It requires the use of the same kind of arguments (sumti) that are used in First Order Language (FOL). To give an idea of what is a proper argument in FOL I quote Barwise & Etchemendy: "Names can be introduced in FOL to refer to anything that can be considered an object. But we construe the notion of 'object' pretty flexibly-to cover anything we can make claims about. We've already seen languages with names for people, sets, numbers, and...blocks. Sometimes we will want to have names for other kinds of 'objects' like days or times." Personally, I think this is stretching FOL a bit. But my point is that there is an existing culture of logicians who write and use FOL. It works. It is sort of the basic English of logic. In the lojban project we have leapt forward to an experimental 2nd order grammar and we are having all the troubles that our model Indo-european languages have. The advantage of having xa'a-xu'u is that we can step back temporarily to a proven, relatively bug-free system, for a word or a book; if we are willing to give up for a moment the vastly increased expressiveness of a 2nd order language. Clarity is gained, expressive power is lost. It is a more restrictive grammar because it does not take predications as arguments. But it does allow us to speak clearly and precisely of objects. Of course we want both worlds, and we can have them. >> >> ABSTRACT SUMTI PLACE======>CONCRETE SUMTI PLACE >> (2nd order) *ge'x (first order) >> >> CONCRETE SUMTI ======>ABSTRACT SUMTI >> (1st order) tu'a (2nd order) >> >> To get djica and her kin working right we can alter the nature of the >> permissible variables with tu'a or alter the nature of the receptor >> site for the variables with *ge'x. A third way is the lujvo route >> suggested by lojbab and .and. JL> >And a fourth way would be to make {djica} mean what that supposed lujvo >would mean, but this is incompatible with the others. > >What I don't like about your proposal is that the meaning of the selbri >is somehow changed by the sumti that fill the places. There are already >many predicates that accept either objects or events, without special >marking. Why should djica &Co be an exception? GK: Djica & her mysterious friends would not be an exception. This device can be used anywhere it would make sense to do so. As to why djica etc. are so defined, I do not know the historical rationale but I am sure there is a good one. I think you are right that there is some change in the meaning of the selbri when an object is in X2 instead of an event. Colin pointed this out to me in the pool example. Also I believe that meaning does not attach neatly to single words but rather to whole sentences and paragraphs. But this is the nature of any language, we can't have an exact selbri for every shade of meaning. JL> >If the answer is because they are often used with opaque references, then >why not mark the opaque references as such, instead of forbidding all >object references? Isn't that throwing the baby with the bathwater or >something? > GK: It doesn't forbid object references, rather it requires them. But only so long as you want: for one sumti, one bridi, or an entire utterence. I don't want to misinterpret you, but just in case you meant to say 'non-object references' here, then I would say that if you wanted to express an event sumti with djica and mark it with some future opaque marker, that is o.k. The grammar shifter cmavo xa'a-xu'u causes transparency through temporary objectification, and transparancy is a side-effect. The main benefit is to be able to speak of objects with selbri that lack an object place where you want it. I also see a use where we want machine-translatable language. JL> >Or do you propose that in cases for example like: > >>> zanru zar zau approve >>> x1 approves of/gives favor to plan/action x2 (object/event) > >we should also use your cmavo when we talk about an object? GK: Yes, if you want to emphasize that X2 is an object. But normally it would not be necessary as anything in LE would be considered an object and anything in NU an event. JL> >There is a simple meaning that can be given to {mi nitcu ti}. Why not >allow it? > GK: Again, I don't know the design history. But under my proposal you *could* say { mi nitcu xa'a ti}. Then "ti" would have to be an object, person, etc. and not an event or a predicate. In any case the xa'a-xu'u method provides a general solution to temporarily re-defining selbri so they can be used in a concrete way as they normally are in first order logic. I havn't seen the ramifications of the change on opacity as I just got this idea a few days ago. >Jorge > -------------- GK: Here is a further example: Cherchez la femme. definition: mukti x1 is the motive for action/event x2 to actor x3 (I take it that both x1 and x2 are event-type sumti.) (2nd order) (current) (event style) (required) ("high lojban") i. tu'a le ninmu pu mukti fi la l'ail l'ovet fe le nu ri ciska le se sanga Something about the woman was the motive to Lyle Lovett for the event of his writing the song. (1st order) (proposed) (object style) (optional) ("low lojban") i. xa'a le ninmu pu mukti fi la l'ail l'ovet fe le se sanga poi ra ciska xu'u The woman was the motive to Lyle Lovett for the song which he wrote. In this case the current, 2nd order expression seems more precise, flexible, and subject to modification or expansion. However the second is preferred for its directness by my son, who writes songs. It is a matter of personal taste. They do connote a difference in point of view: one rooted in objects and the other in sets or classes. Both these sentences are, well, understandable English. The current parser sees both these sentences as error-free. Why not allow both these sentences? Here is some background from the person who laid the blueprint for this language, lojban, at least in my opinion, Quine: On 2nd order language: Thus if we extend truth-function theory by introducing quantifiers '(p)' '(q)', '(Ep)', etc. we can then no longer dismiss statement letters as schematic. Instead we must view them as variables taking appropriate entities as values, namely, propositions or, better, truth values, .... We come out with a theory involving universals, or anyway abstract entities." "The ontologically crucial step of positing a universe of classes or other abstract entities can be made to seem a small step, rather naturally taken, if represented as a mere matter of letting erstwhile schematic letters creep into quantifiers. Thus it was that '(p)' was admitted unchanged into quantifers ..... Similarly, in an imaginative reenactment of the genesis of class theory, let us now consider in detail how class theory proceeds from quantification theory by binding erstwhile schematic predicate letters." "Classical mathematics has roughly the above theory as its foundation, subject, however, to one or another arbitrary restriction, of such kind as to restore consistency without disturbing Cantor's result." "By treating predicate letters as variables of quantification we precipitated a torrent of universals against which intuition is powerless. We can no longer see what we are doing, nor where the flood is carrying us. Our precautions against contradictions are ad hoc devices, justified only in that, or in so far as, they seem to work." On 1st order language as an only language: "The heroic or quixotic position is that of the nominalist, who foreswears quantification over universals, for example, classes, altogether. He remains free to accept the logic of truth functions and quantification and identity, and also any fixed predicates he likes which apply to particulars ... He can accept laws which contain variables for classes and relations and numbers, as long as the laws are asserted as holding for all values of those variables; for he can treat such laws as schemata, like the laws of truth functions and quantification. But bound variables for classes or relations or numbers, if they occur in existential quantifiers or in universal quantifiers within subordinate clauses, must be renounced by the nominalist in all contexts in which he cannot explain them away by paraphrase. He must renounce them when he needs them." Try xa'a, you'll like it. djer