[YG Conlang Archives] > [engelang group] > messages [Date Index] [Thread Index] >
pycyn@a... wrote: > This comparing apples and pears -- almost the same thing but > not quite. the parse is a strictly linear and binary breaking > of the sentence to determine whether it is grammatical and how > it is constructed according the this particular grammar (the > official one). This is meant to be very close to the syntactic > structure of the sentence but pretty obviously is not identical > with it. The difference is often, as here, that the > synstructure is not binary, so the transition is simply to > ignore certain kinds of divisions that the parser introduces > (there are several other artifacts of the parser, whose charm > is that it has all sorts of nice technical features and is > relatively easy to work with, and some of them are much worse: > sentence modifiers that get bound to the nearest word, for > example. This may be what & is talking about when he says it > is a long way from being a logical language). I want to make sure I grasp this. Are you saying that the parse reflects the official grammar, but this is not identical (though close) to the syntactical structure, which can only be obtained by knowing what divisions are to be removed from the parse? In short, I perceive three levels of analysis: 1. The parse, which determines grammaticality (legal strings of words). 2. The syntactical structure (most of the meaning, but not things like scope), which can be obtained from the parse by removing some of the divisions. 3. The logical structure, which can be obtained from the syntactical structure and the original utterance by various "other means", most notably by inspecting word order. This sums up what I am reading. I hope you will forgive me if I misunderstand you. > I don't, incidentally agree that the predicate must govern its > arguments, though I admit it is hand for a lot of things. But > handiness, rather than theoretical elegance is its main virtue. > I actually think that the Lo*an parses are better for logic > (FreeMod et simul aside) than many putative linguistic tricks. I think it's moot whether you choose the strict predicate- argument structure or the centuries-old subject-predicate structure. There may well turn out to be an arbitrary number of parse/production rule sets that can be defined for any single given grammar. My own intuition would guide me towards a parse that at least captures the syntactical structure of the words, or, alternately, the logical structure, or, if possible, both. I admit there may well be issues here I am not aware of, at least specifically to Lojban. > But, then, I am in this area yet another clique of one. I tend > to think that in the case of this instant sentence it is almost > perfect: I would have done it as ro(nanmu (cu prami (lo > ninmu))), an A-proposition with the main internal break > between subject and predicate (of course) and a complex > predicate (in the older sense of those terms). I would not > expect the parser to give me this, but surely to give me -- as > I think it does -- good input to get this. I agree it's pretty close to the logical structure. Close enough to have me wondering to some extent why it wasn't made to match. Incidentally, are you familiar with lambda calculus? I am revisiting this concept from my studies of LISP/Scheme. Along with their "lists", as I mentioned to And in another posts, I think that these things might possibly be applied profitably to the parsing of loglangs. Regards --- Mike