* Wednesday, 2014-10-22 at 20:38 -0300 - Jorge Llambías <jjllambias@gmail.com>:
> On Wed, Oct 22, 2014 at 8:08 PM, Martin Bays <mbays@sdf.org> wrote:
>
> >
> > ba gi broda gi zajba == broda .i ba bo zajba
> > bai gi broda gi zajba == zajba .i bai bo broda
> >
> > Which reminds me - which way should
> > broda .i ba je bai bo zajba
> > work, or should it be some mix? If a mix, how about with {joi}?
>
> I guess it would be: "ba je se bai gi broda gi zajba" and the same for
> "joi", but we should unify all tags to work the same way.
That would be nice!
I'm not sure that using {se} like this is robust, though. There's no
formal reason that
[BAI] gi bakni gi zajba
should be equivalent to
se [BAI] gi zajba gi bakni
, even if it often is, right?
> > > I think it might also make sense to analyse tag-connectives as if
> > > they were ordinary tags on the second connectand, with the first
> > > connectand acting as their complement, so:
> > >
> > > broda .i [jek tag] bo brode
> > > -> broda .i [jek] brode [(se} tag] lo nu broda
> >
> > But if broda doesn't occur, what is {lo nu broda} referring to here?
>
> In that case the complement would have to be "lo nu na broda". So:
>
> broda .i [jek tag] bo brode
> -> broda .i [jek] brode [(se} tag] lo nu xu kau broda
>
> where "lo nu xu kau broda" is whichever of "lo nu broda" and "lo nu na
> broda" is true.
I don't immediately see exactly what that would mean, so let me back up
and think abstractly about events, tenses and modal operators.
A lojban statement is interpreted as a proposition in the corresponding
logic, and by default the illocutionary force is to assert the truth of
the proposition. Often the assertion is spatiotemporally restricted, and
we can say that an event (of the proposition being true) is being
asserted to occur. Again, for this discussion I'll pretend (or note?)
that this is the only case to consider.
In lojban, and I suppose in human cognition, events are things. They can
be counted and measured in various ways, and can satisfy predicates.
A tag is interpreted as mapping an optional term to a modal operator
- something which takes formulae to formulae. The semantics of these
operators are highly miscellaneous; {ba}, {no roi}, {ja'e}, {bau} all
have quite distinct flavours.
Now to return to tag connectives, and to correct some mistakes I made
earlier in this thread, let me first think through again the simple case
of {[tag] gi broda gi brodu}.
Earlier I implied that this could be handled symmetrically, but that was
just wrong-headed; e.g. in {no roi gi broda gi brodu}, we're claiming an
event of broda occurs, but not so for brodu.
So it must be something like
"an event E of broda occurs; {[tag] E brodu}".
I think it must actually be:
[tag] gi broda gi brodu
<=> su'o da tu'e da fasnu gi'e nu broda .i [tag] da brodu
, or maybe that but with {su'oi} in place of {su'o}.
(For this to make sense, {fasnu} must behave the with respect to
tense the way "occurs" does, e.g. an event which occurred in the past
fasnued then but doesn't fasnu now.)
Evidence for this being the right thing: we can rule out any "lo-style"
semantics, where some definite nu broda are involved, by considering how
the construction works in negated scope:
na ba gi broda gi brodu
should make sense, and be true, if no events of broda occur. So {lo ge
nu broda gi fasnu} can't be involved, nor is it about a specific
contextually specified nu broda, nor the kind, nor anything similar that
I can think of.
Universal quantification would be silly. So by elimination of the other
natural choices, it has to be existential quantification.
In particular, {broda .i brodu [tag] lo nu broda} may be a handy
shorthand reformulation, but it isn't actually equivalent.
Now:
> broda .i [jek tag] bo brode
> -> broda .i [jek] brode [(se} tag] lo nu xu kau broda
. For simplicity, let's assume the tag is a tense (so there's no {se}).
Again, if this is to work in embedded scopes, I think we need to recast
it in terms of quantification over events.
Annoyingly, I can't see a way to do that that's uniform over the
possible jeks. But e.g. for {jo}, I guess it would have to be:
da zo'u ge da fasnu
gi ga ge da nu broda gi [tag] da brode
gi ge da nu na broda gi [tag] da na brode
Is that faithful to your intention? Is there a way to abbreviate it
using {xu kau}? I don't really know how this kind of use of q-kau works,
but I'm guessing it jumps scope like actual questions do, so can't
really be used here for this kind of thing.
Anyway, this does seem like a reasonable semantics.
To compare with the TT-skimming approach I suggested:
{broda .i na ja ba brode} still means "if broda then then brode";
{do ba prije gi'o ja'e bo snada} adds that failure will be caused by
non-wisdom; in the CLL example
{ba tu'e mi bevri le gerku .ijacabo mi bevri le mlatu}
the truth conditions seem to be the same without the {ca}, but then
I think that's how it has to be read in CLL too (so it isn't a very good
example, really.), and this is also the case with TT-skimming.
Martin
Attachment:
signature.asc
Description: Digital signature