X-Digest-Num: 327 Message-ID: <44114.327.1776.959273825@eGroups.com> Date: Mon, 3 Jan 2000 05:02:55 EST From: Pycyn@aol.com Subject: Some peripheral notes on Legalese Lojban X-Yahoo-Message-Num: 1776 Content-Length: 3497 Lines: 55 Actually not on the legalese part at all, but on some remarks that were made in the discussion. 1) Just Predicate Calculus. As & showed in his own examples, this is just not enough. Even patents (maybe especially) need a) second order predicates (take predicates or propositions as arguments) and probably second order quantifiers (though there are technical tricks around both of these which only occasionally screw up an inference) and b) intensional contexts (hypothetical situations, for a minimal example). 2) This realization -- late 50's and early 60's -- led JCB to move from his original program to a richer language (though one without -- thank God -- all the typographical stuff of L1960). This was still within his SW project, since he came to see that he had to have alnaguage the speaker could fully inhabit, not merely use for a few hours in an artificial project situation, to give the hypothesis a real test. Many of the features that were added grew out of his own and others' efforts to inhabit the language for real life situations. Even the small number of rather limited efforts in that direction (a few dozen stretches of an hour or so in Loglan only, some attempts to teach various infants in Loglan) pointed to still further needs. Even more came from attempts to translate both literature and ordinary texts -- though many of the innovations first sought for these proved to be unnecessary, resulting from inadequate creativity in using available resources (and from incredibly odd choices in what to try to translate for starters). A second source of innovation was heuristics, what was needed to teach the language effectively. The Great Morphological Revolution, which mainly made all compounds uniquely decomposible, was one such, as was the effort to find a uniquely parsible grammar -- and thus provide at-home checkers for students. (I suspect that JCB once thought that the decomposition algortihm for predcalc would simply carry over to Loglan, no matter how many furbelows were added.) Clearly, a lot of this stuff is not needed in patent applications (most attitudinals, for example, but not all, are irrelevant). But, since just about everything is optional in Lojban, patent attorneys could take what they want and leave the rest, while someone else might find a yet for some of the rejected peices for some other international area. As for Lojban as an intermediary language in machine translation, etc., my experience was that it was harder for the linguists of that time (1960-2) to write English to logical English programs (ditto Russian, and reverse) than to write English to Russian (or conversely) programs -- not that any of them of any sort were very good. Nowadays, of course, more linguists know quite a bit about predicate calculus and the like (I had my linguistics professors in my early logic classes at the beginning of that wave) and all the various possibilities have been expanded quite a bit (ProtoThinker, as I have mentioned before, has a serviceable English to logical English rewriter and the reverse and is very small). I suspect that, given the incentive (e.g., a body of material to be turned into Lojban and disseminated in a variety of languages), reasonably good (i.e., as good as any language-language machine translation) programs would be forthcoming in a very short time. A good human editor still would be required at each step, alas. pc