[YG Conlang Archives] > [engelang group] > messages [Date Index] [Thread Index] >
Mike S., On 30/08/2012 16:25:
On Thu, Aug 30, 2012 at 10:25 AM, And Rosta <and.rosta@hidden.email <mailto:and.rosta@hidden.email>> wrote:
I would approach the rules differently, by defining the combinatorial properties of words, so the thing currently called "complement" wouldn't arise. I would normally have said a unary operator has one complement and a binary operator has two complements, but one could call them "dependents" instead, to avoid confusion.
Would "formula-modifier" be acceptable to you? Then we can use
"complement" in the manner that you want it to be used, to refer to
formulas governed by operators. I would prefer *not* to change the
rules, but I agree that "complement" as it is used now has a slight
smell.
OK. I see those rules as a kind of informal backoftheenvelopey sort of description, so I'm not that fussed about them, but yes, it's good not to take technical terms from linguistics and then use them in a different incompatible technical sense here.
However, I do object to using "dependent" to refer to formulas, i.e. _well formed formulas_, which is exactly what they are. Xorban formulas are *not at all* analogous to natural languages "dependents" -- they are analogous to sentences and the free-standing heads of sentences. Unary operators and BR-expressions (binder-restriction expressions) are the real dependents, precisely because they are not well formed formulas. Unlike formulas, but like like natlang adverbs and noun phrases, they *cannot* stand alone. The notion that formulas are the dependents is simply backwards.
I'll explain why I think you can't be right. 1. Semantic composition. A dependent's meaning is complete in itself. A head's meaning is incomplete until the dependents are added. Ergo "ja" is head with two dependents, "na" is head with one dependent, "r-" is head with two dependents. 2. A head may have multiple dependents but a dependent may not have multiple heads. Ergo in "ja/ra X Y", ja/ra is head of X and Y, not dependent.
Basically, I'm influenced by the way natlangs work, on the grounds that ideally the loglang would be usable as if it were a natlang. But as with phonology, the syntactic description can come later.
Xorban is pretty different from natural languages, so the terminology
is going to be at least a little different.
The grammatical categories will be different -- no need for nouns, verbs, subjects, objects and so forth. But the basic mechanics are the same -- at least, I would stipulate that as a fundamental design goal, else why insist that the language have a linear phonetically-realizable phonology, for example. I'm pretty sure that we all [of us taking an interest in Xorban] take this design goal pretty much for granted, tho if it turns out we don't then that would make for an interesting discussion.
To the extent that it patterns like a natural language, it very clearly patterns like a rigid SOV language in which the S and the O are BR-expressions and the V is a formula.
There is no S and O. That is, the verbs/predicates have inflections marking subject and object, but there are no separate subject and object phrases. Well, I can see how you could see it your way. You could see "la xrma mrsa" as having the structure of "[[the [horse]] died]", "[[la [xrma]] mrsa]", rather than "[la [xrma] [mrsa]]". But I can't see a rationale for that, and I can see obvious problems with it. --And.