On 15 October 2015 at 03:31, J S Jones <[log in to unmask]> wrote: > On Wed, 14 Oct 2015 22:40:59 -0600, Logan Kearsley <[log in to unmask]> wrote: [...] >>I spent more time than I should have puzzling over the orthography >>section; I think it would help to change "Each symbol is composed of 3 >>components: 1 from each line, representing the octal digits." to say >>"...one from each row..." instead. I was thinking of the lines that >>make up the glyphs, and was thoroughly confused. > > I guess "line" should be "row of table"? That would be even clearer, yes. :) >>In Derivational Morphology, the Concatenation section has "Insert math >>here"; I would like to see that math! It is not clear to me exactly >>what "product" or "composition" mean in this case. > > I'll need to look it up in the compiler theory textbook, but the idea is > > R3 (a, c) = R1 (a, b) x R2 (b, c) > > where R1, R2, and R3 are relations. The thing that isn't clear to me is why that doesn't form a trivalent predicate. I suppose the 'b' argument just gets discarded? That would fit with your kinship term usage- if, say, "father-father" meant "paternal grandfather", and you don't care about mentioning the father in the middle of that relationship. In which case the math, if you even need to include it (which I'm not certain you really do) would be a join of the two relations on the shared argument (or, if we want to get really strict, a rename so that X, Y columns of each predicate are renamed to the appropriate a, b, and c labels, followed by a join) followed by a projection onto the remaining a, c / X, Y columns. >>Also, I'm a little unclear on how you figure out which arguments are >>which when there's a mixture of full phrases and index clitics. > > Where there are 2 arguments, the full phrase matches the non-indexed argument, if that's your question. I would think that the instances where all arguments are phrases would be more difficult. Or is your question about how to know the argument structure of a given predicate? In > > child=Def=3C pie=Ind eat-Prf=Tmp fork=Def 3C-wash=Fin > > the main predicate "wash" has a prefixed agent (3C) and a suffixed patient (0). I kinda figured if all the arguments are phrases, then they just go in X-Y-Z order. In this case, I'm guessing the 0-patient suffix counts as the "non-indexed" argument? If so, then I can see how that sentence works, but how do you, for example distinguish between an indexed Y with a 0 Z prefix, vs. a 0 Y and indexed Z? Is that just handled by what semantically "makes sense"? -l.