* Saturday, 2011-11-05 at 22:28 -0300 - Jorge Llambías <jjllambias@gmail.com>:
> On Sat, Nov 5, 2011 at 8:34 PM, Martin Bays <mbays@sdf.org> wrote:
> > * Saturday, 2011-11-05 at 18:18 -0300 - Jorge Llambías <jjllambias@gmail.com>:
> >
> >> I think I do get it. I just don't think it has anything to do with
> >> logical structure.
> >
> > Well that's a matter of definitions.
> >
> > But note e.g. that the classic example of scope ambiguity in english,
> > "someone loves everyone", can be looked at this way:
> >
> > A: "Someone loves everyone."
> > B: "Oh yeah? Who?
> > A: "Their mother."
> >
> > A: {su'o prenu cu prami ro prenu}
> > B: {ma prami ro prenu}
> > A: {lo mamta}
> >
> > (Lojban can't seem to get at the "their" in "their mother", but that's
> > not really important)
> >
> > (and yes, I know by now that you would consider A to be breaking your
> > favoured domain conventions by having both mundane people and Mother as
> > a person in the same domain; but (a) that's an informal rule, which
> > appears to be flexible (you broke it in the xabju example), and (b) it's
> > not important to the essence of the example that prenu is being used on
> > both sides)
>
> I still don't think that's a matter of logical structure. It's A
> tricking B into one interpretation to get an effect once the "right"
> interpretation is presented. That's how many jokes work.
Well, I presented it in joke form - which was possibly foolish as
I didn't intend to trivialise the issue!
Really, I don't see that the situation is significantly better than it
is in english.
A search for "quantifier scope ambiguity examples" yields various
examples of the issue in english, most of which appear to go through
directly in kindful lojban.
Another clear example:
"A professor talked to all the students"
{su'o ctuca cu tavla ro le tadni}
could mean only that each student was talked to by a professor -
formally, just because the kind Professor ctucas; or if we apply your
informal rule that quantification indicates that there should be
multiple things at the same level involved, then because it could be
that they were all talked to by a logic professor.
> >> Consider "a beret is a type of hat". I would say "lo ranmapku cu klesi
> >> lo mapku".
> >
> > In reality, I'd just say {ro ranmapku cu mapku}.
>
> What about "berets and bowler hats are different types of hats"?
> "lo ranmapku jo'u lo bolmapku cu ficysi'u lo ka klesi lo mapku"
Again we could avoid kinds, and just say {su'o da ranmapku .o nai
bolmapku}. Or we could use properties rather than kinds, and say {lo ka
ranmapku na du lo ka bolmapku}, or copy your approach with {lo ka
ranmapku ku jo'u lo ka bolmapku cu ficysi'u lo ka kairni'i lo ka mapku}
(where ro da poi selkai ku'o ro de poi selkai zo'u go da de kairni'i gi
ro di ckaji da na.a de) (although {go'e fi lo ka ma kau ckaji} might
make more sense).
> > But if you forced me to use kind terminology, I'd want a second
> > predicate for "x1 is a subkind of x2". From the gimste definitions, I'd
> > be more likely to use {klesi} for that than "x1 is an instance of x2",
> > which is closer to {mupli}. In fact, {mupli} seems to want a property in
> > x2, so maybe this could be {klemupli}.
>
> (I would rather re-define "mupli" into "x1 is an instance of x2", but anyway.)
>
> ...
> > But maybe it's true that kinds are useful enough that the language
> > should have special facilities for handling them - e.g. allowing {lo
> > mapku} to get a kind. We just need to have ways to disambiguate.
>
> "klesi" allows us to disambiguate between two levels. Disambiguating
> between a potentially infinite number of levels is trickier. As the
> old Lojban saying goes: the price of infinite precision is infinite
> verbosity
Can you give an example where we might want to go up two levels from
mundanes (as opposed to their stages or whatever)? I wouldn't be
surprised if there were such, and maybe you've given examples before,
but none spring to mind (other than artificial examples like "kinds of
kinds of garment" - unless you can think of natural cases where we'd
want to talk about those).
> > The "imaginaries" terminology of the other thread gives one plausible
> > approach to this - treating kinds as analogous (and, in a sense, dual)
> > to bunches. {su'o} would get neither bunches nor imaginaries, but {lo}
> > could get either.
> >
> > I suspect that a system based on this could explain e.g. most if not all
> > of the sentences in your alis, while also being sufficiently
> > disambiguable to satisfy me.
> >
> > Would you reject such a solution out of hand?
>
> I think that covers most needs, but I suspect there are cases when we
> may want to quantify over kinds.
Hmm. That didn't sound like a rejection!
For quantifying over kinds: if the rule is that {lo} gets a bunch of
imaginaries which are all imaginaries with respect to the same
equivalence relation aka differentiation criterion (i.e., to import one
more piece of model theoretic parlance, a bunch of imaginaries from the
same "imaginary sort"), I see nothing wrong with using e.g.
{ca lo prulamnicte mi citka vo lo cidja poi do nelci}.
I would also want it to be possible to specify that we are fa'u are not
talking about imaginaries (with respect to a non-trivial equivalence
relation, i.e. one coarser than equality), perhaps with {lio} fa'u
{loi}.
(No that wasn't a typo! The PEG morphology allows {lio} as a cmavo form,
right?)
I'd also want to be able to specify the equivalence relation in question
in the former case, i.e. as per And's (iii) of the other thread. I don't
know how to do that... maybe with inner quantifiers?
{re lo fi'u vei ni'e ka skari ma kau ve'o mapku cu vi zvati} for
"two colours of hat are here", or
{so'o lo fi'u vei ni'e ka danlu ma kau ve'o cinfo ba zi morsi} for
"several species of lion will soon become extinct"?
With {lio broda} being (blissfully) short for {lo fi'u vei ni'e co'e ve'o
broda}?
And {lo fi'u ro cinfo} being the wholly singularised lion, i.e. Lion
(rather than an infinitesimal amount of lion)?
Martin
Attachment:
pgpjwVpWvguL5.pgp
Description: PGP signature