* Sunday, 2014-09-28 at 20:17 +0100 - And Rosta <and.rosta@gmail.com>:
> On 28 Sep 2014 17:03, "Martin Bays" <mbays@sdf.org> wrote:
> > * Sunday, 2014-09-28 at 12:35 +0100 - And Rosta <and.rosta@gmail.com>:
> > > On 28 Sep 2014 12:04,and.rosta@gmail.comwrote:
> > > > On 28 Sep 2014 02:34, "Martin Bays" <mbays@sdf.org> wrote:
> > > > > the compositionality of restrictive clauses (which admittedly is
> > > > > already broken in some other cases, e.g. {da poi broda}).
> > > >
> > > > I agree {da poi} looks broken. What are the remedies? (1) To
> > > > allow noncompositional idioms? (2) To define /poi/ as an
> > > > allomorph of /noi/ in this syntactic environment? (3) To accept
> > > > that, given the internal logic of the language, {da poi} as
> > > > habitually used is simply wrong? (4) To seek and find
> > > > a consistent definition for {poi} and {noi} such that {da poi}
> > > > usage becomes correct?
> > (5) change the grammar to match the semantics, taking variable out of
> > sumti6, and having restrictive relatives allowed only where there's
> > a matching place for a quantifier;
> > (6) note that doing (5) wouldn't actually render anything ungrammatical,
> > so treat the existing grammar as a kind of shorthand for that in (5),
> > and so consider current handling to be "effectively compositional".
>
> If {X poi broda cu brode} = {lo me X gi'e broda cu brode} and {X noi broda
> cu brode} = {X broda gi'e ...... brode}, then neither strike me as
> appropriate for restricted quantification.
Agreed.
> To get at the underlying logic
> of {so'e broda cu brode}, it seems to me you need something like {lo du'u
> ke'a broda kei so'e zei so'e lo du'u ke'a brode}, where {so'e zei so'e} is
> an ad hoc selbri counterpart of so'e, and which xorxes would surely want to
> simplify to {lo broda cu so'e zei so'e lo brode}. (This is similar to but
> less adequate than the solution I got introduced into Xorban.)
>
> So if there is a compositional interpretation for {da poi}, can you gently
> and not too technically spell it out for me?
Sorry, I wasn't trying to claim one in that sense. All I meant was that
we can consider the semantics to be compositional where one of the
components is "quantifier? variable restrictive-relatives?", and another
separate component is "quantifier? sumti6 restrictive-relatives?", where
we give these the (different) semantics discussed.
In other words, I proposed cheating!
> > > Modifying a variable, that cannot be the difference. I had been
> > > assuming that noi modifying a variable would have outermost scopal
> > > position within the domain in which the variable is bound.
> >
> > Hmm, that makes some sense. I have been taking, ever since xorxes told
> > me off for doing otherwise in the first release, incidentals to be
> > truly incidental, having no effect on the truth value of the main
> > proposition. So a {noi} on a variable yields a side claim entirely
> > outside the scope of the corresponding quantifier, so involving an
> > unbound variable, which I'm currently (fairly arbitrarily) handling by
> > universally quantifying it out over whatever domain it was originally
> > quantified over, so e.g. {du su'o da poi broda zi'e noi brode} -> {ro da
> > poi broda zo'u da brode .i su'o da poi broda zo'u du da}.
>
> It seems to me that my suggestion better yields a single rule that applies
> consistently to all cases. E.g. {ro broda noi brodo na brode} = {ro broda
> ku brodo gi'e na brode}.
So e.g. {naku ro broda noi brodo na brode} = {naku ro broda ku brodo
gi'e na brode} == {su'o broda ga na brodo gi brode}?
Having fully scope-breaking incidentals seems more useful.
Martin
Attachment:
signature.asc
Description: Digital signature