[YG Conlang Archives] > [engelang group] > messages [Date Index] [Thread Index] >
Mike S., On 28/08/2012 03:43:
Brevity must be the guiding principle.
A logical language is one that encodes logical structure (predicate--argument, operator--variable structure) -- preferably unambiguously.
Brevity and logic are orthogonal principles;
That's right. But the logic principle is easy to satisfy; the brevity principle is hard to satisfy. Or rather, brevity is an aspect of a more general principle of usability -- being better than natlangs in some respects (unambiguousness) and no worse than natlangs in others (usability).
IMO what's important is that "do'a xrbno'a je je lgjo'a bngo'a nu je stro'a bngo'a".
What are d-, nu & str-?
> How about "sa sma se sme prmake"?
>
> prmo'e.
How come? Are you sure?
I think so. I don't like the idea that dropping a variable changes a
predicate, and the most obvious interpretation rule for dropped
variables is o'e.
Ah, okay. This came up in an earlier reply. We're diametrically opposed on this one. Well, my preference would be for a system that allowed free elidability of places, but that system can't be based on the current morphology. An alternative system (much more like Livagian) would be to mark the argument-place by the vowel type (e.g. "love" would be prmi'u & prmu'i) and then order the vowels so that they are syntactically bound by nested binder--bindee relations. A huge virtue of this system is that the (fluent) speaker needn't memorize variables' names, because variables have no names, being purely syntactic with no morphological form. The downside is that the rules for binder--bindee relations are likely to be either too restrictive or very complicated or require a load of extra inflectional marking. But this would be a huge change from the basic workings of Xorban.
> > All of that aside, would provide any method way to create a sort of
> > anaphor that would be represented by free variables? If so, how would
> > it work?
>
> I don't see a need for free variables. Variables not explicitly bound can be implicitly bound or can be of the abbreviatory a'a sort.
>
>
> (1) na sa xrja vfla. na se sme [la sma] nlveka.
> "Pigs don't fly. Something has no wings."
>
> I don't think it'd be long before humans started hearing that as:
>
> (2) na sa xrja vfla. na se sme [la xrja] nlveka.
> "Pigs don't fly. They have no wings."
>
> ... and then started hearing [la Ra] everywhere. Do you disagree?
I don't disagree, tho (2) is unlikely given the availability of the set of variables that are interpreted as definites when unbound.
Okay, I will have to give them a look. There has been a burst of
activity and it's hard to keep up.
I can't keep up either, and my holiday isn't even over yet, though it will be soon, which will probably cripple my involvement in these discussions.
> As for anaphors, I'd use a pair of predicates meaning "co'e", one with and one without a description argument as a syntactic complement. E.g. "xx-" is "co'e", "is it", and "Xz" has the extra complement, "xza qa grka" (or indifferently, "xza qo grko", "is the dog", where "q" is the ce'u quantifier and "qa grka" means "the property of being canine". maybe you wouldn't consider these anaphors, in which case I'm saying I don't see the need for anaphors.
>
> If "xza qa grka" means is "is the dog", does that mean a specific
> dog, or does it mean dogs in general?
A specific dog, tho that specific dog could be the generic dog.
Well then I don't see the difference from "la grka". Nothing excludes
a specific reading with "l-".
If l- is equivalent to q-, as Jorge suggested, then l- does exclude a specific reading. But anyway, xx- and xz- encode that there definitely is a "co'e" element (and maybe also that the description is nonveridical).
> There's also the possibility of using stems like qam- (= 'am-), where qa- introduces vowelless name stems, and the name is taken to refer to something already referred to with a predicate starting with m- (ignoring any name-introducing prefix), like Lojban my.
>
> That was always a nice scheme, but I think that Xorban's explicitly
> bound variables tend to obviate the need for that.
But this scheme is used for predicates or referents that are repeated across sentences.
What makes this scheme better than assigning implicit restrictions to
free variables across sentences? Either way, you need some variety of
discourse representation theory to formally represent the persistence
of the restrictive information across sentences,
Explain further. I don't see this need, but probably I don't understand well enough what you mean.
and I say using the variables and more at home in Xorban than using a novel morphological class of Cy-like entities. Plus, variables are actually explicitly bound to their restrictions at least once; these other things are glorked.
As Xorban stands, you don't have to understand it as involving restrictions on variables, and I like that. That is, in "l/s/ra bcda fgha", "bcda" needn't be understood as a restriction. Of course it's possible that this is merely terminological: if definites have antecedents, the antecedents will (in Xorban) be properties, and the dependents of l/s/r can be understood as encoding properties, and then all your rule is doing is privileging the first dependent and associating the property it denotes with vowel used for the variable in it. Still, even then I dislike privileging the first dependent (because of the arbitrariness and the lack of a way to target the second dependent) and obliging interlocutors to make a mental note of the vowel associated with each first dependent (because of the big extra mostly-superfluous load on working memory). --And.