[YG Conlang Archives] > [engelang group] > messages [Date Index] [Thread Index] >
Mike S., On 29/08/2012 21:40:
On Wed, Aug 29, 2012 at 11:03 AM, And Rosta <and.rosta@hidden.email <mailto:and.rosta@hidden.email>> wrote:
Mike S., On 28/08/2012 02:33:
> Glossing over the quotatives for a moment, that would make {la
> "djan"a} mean (in Lojban) {zo .djan.} instead of {la .djan.}. To get
> {la .djan.} you'd need {la "djan"ako'e}. Surely the latter, which
> we'll need much more often, should have the shorter form.*
Fine. I thought that by saying "Y (is known as X to) Z" you were intending that the suffixes should by VkV always. I agree a plain V suffix is warranted for brevity. I also think my suggestion was independently poor because it wouldn't allow for the quoted string to contain /q/.
I was thinking that this particular ("Class I") quotative would be
used for nouns, both common and proper, non-fussily transliterated
from various natural languages into a string of Xorban segments. The
transliteration could not contain Xorban's own <q>=/?/, but neither
could it contain English's /T/ nor French's /y/.
You're assuming here for the sake of exposition that Xorban has Lojban's phonemes, right?
I was thinking that one could use a more powerful but more
complicated ("Class II") quotative for quoting a broader range
phonological material.
I *think* the broader range of phonological material comprises only /q/, since everything else can already occur within q-delimited quotes. Am I missing something?
> The need for referencing the name is less urgent than in Lojban,
> because people can say {"djan"a'a} ="I am John" or {la prmaka'a
> "djani"a'aka}="I am 'Johnny' to my friends." Of course we do need to
> reference the name. How about:
>
> CV1+ X + q = "x"(y, z) which means "Y (is known as X ending in vowel to) Z
> CV2+ X + q = "x"(y, z) which means "Y (is known as X ending in cons. to) Z
I'm not sure it's worth it. Ambiguity over whether the last vowel was included would occur only with /@/, and that ambiguity is livable with.
I thought of exactly this myself yesterday! I think the rule for /@/
is like that of /?/ in "Class I" names; /@/ is reserved for Xorban's
internal uses and therefore is not a segment available to be
transliterated into. English "Jessica" = /'djE.s1.k@/ might then be
transliterated <CVdjesikaq-> where the English /@/ is transliterated
as /a/ to indicate the original final vowel, and Xorban /@/ remains
free to buffer /dj/. "Mark" would be transliterated <CVmarkq-> -- no
original final vowel, no original vowel between /rk/ either.
I agree.
> *Moreover, I think that all elided trailing variables are implicitly
> "o'e",
I'm not too happy with that. First of all, it requires a rule that multiple unbound o'e are independently bound, i.e. o'eko'e is not reflexive, and I'm not sure such a rule would be desirable.
Right, the idea that I had was that "o'e" was equivalent to an elided
argument, and vice versa, and that each underlying occurrence of
"Po'e" would be interpreted as "lX smX PX" where X was a unique
anonymous variable -- I would introduce X just in case "la sma" by
itself wasn't vague enough, as there is still confusion about the
exact semantics of "l-" and "l- sm-.
I withdraw that first objection of mine.
Second, it overprivileges the ordinal priority of suffixes, and creates competing imperatives for ordinal priority -- is ordering done by semantic role or by frequency of being o'e-bound and hence elidable?
I am afraid I don't follow.
I'm assuming you (even tho Jorge wasn't) were envisaging a system whereby if tvl- is defined as X tells Y Z, it can be tvlekaku "E tells A U" or tvleka "E tells A (something)" or tvle "E tells someone something". Given this sort of system, one would want the arguments ordered by how frequently one would wish to o'e them, the least often o'e-ed one first and the most often o'e-ed one last: that would maximize brevity. But that's a tricky calculation, and conflicts with the idea of trying to order arguments by their semantic role. In my system, if, say, there's a ternary predicate one of whose arguments will often be o'e, then we can just add to the lexicon a binary predicate with the same same but without the o'e-prone argument.
Perhaps it would help to explain why "a'a" doesn't create the same sort of problem.
It's my turn to not understand... In what way could a'a create the same sort of problem?
Rather, each stem should have its definition relative to its number of arguments. PtkV and PtkVkV are different predicates, tho it may be that PtkV happens to be defined as an abbreviation of PtkVkV.
I am afraid I can't imagine a situation in which it would be helpful to allow PtkV not to be defined as an abbreviation of PtkVkV; it seems a disadvantage to arbitrarily allow predicates to have more than one lexical entry.
it wouldn't be a correspondence between one predicate and many lexical entries, it would be a correspondence between one stem and many predicates, each of which has one lexical entry. The arity of a predicate is a defining property of the predicate, as you of course are well aware. Allowing one stem to be recycled for different predicates would done not arbitrarily but on principle, the principle of respecting the fact that a predicate is partly defined by its arity, and the principle of maximizing brevity (by optimizing assignment of shortest forms to frequentest meanings). --And.