Re: [lojban] Re: ralju bangu be le gligu'e

[John E Clifford <clifford-j@sbcglobal.net>]

--- Jorge Llambías <jjllambias@gmail.com> wrote:

> On 5/6/06, John E Clifford
> <clifford-j@sbcglobal.net> wrote:
> > Lojban parses, in other
> > words, do  not give relaiable information
> about
> > the structure of the utterance; at best they
> > accept all and only legitimate utterances of
> > Lojban for some different reasons.
> 
> I agree that in Lojban sometimes (not most
> times, but yes in a few
> cases, sometimes inexplicably) the form of a
> sentence, as given by
> the grammar, does not match the meaning as
> given by the interpretive
> convention. I don't think this applies to the
> present case however.

As I note, the misrepresentation (under simplest
interpretation)  begins aat the most basic level,
with the analysis of simple bridi.
 
> > Now to the case in point. {ralju bangu be le
> > gligu'e}.  It is a bridi minus one argument
> 
> It is a selbri, not generally a bridi. Any
> selbri can by
> itself constitute a bridi, so this selbri in
> particular could
> be used as a bridi, but in general it is just a
> selbri, and
> especially in this case, as you say:

This is what I mean about getting off the point
and talking past eachother.  To be sure, the
expression is a selbri, but it is one because it
is a bridi less one argument (and with an
incidental arguemnt marker).  That is, the
interpretation of it is intimately tied to the
interpetation of the bridi made of this plus the
missing argument.  To insist that the logic
involved here is somehow different because this
is  a part of a sumti rather than a bridi is to
misunderstand a major portion of the way logic
works.
 
> > (I
> > assume this is pulled out of {lo ...} or so),
> 
> And what {lo} does is convert a _selbri_ into a
> sumti. It needs
> its input to have empty slots. {lo} cannot take
> a bridi to convert
> into a sumti. In Lojban, the way to convert a
> bridi into a sumti
> is by first converting it into a selbri, with a
> member of NU, and
> only then into a sumti with a member of LE.
> 
> > so
> > it ought to divide [ralj bangu] (the
> > predicate/selbri) and [- (be) le gligu'e]
> (the
> > arguments)(the dash is for the term bound up
> in
> > {lo} or whatever).
> 
> The whole selbri is bound up in {lo}. The full
> structure with all
> terminators included is:
> 
>   lo [ralju (bangu be le gligu'e ku be'o)] ku
> 
> {lo} takes a selbri (in this case one
> consisting of a two component tanru)
> and converts it into a sumti. Notice that {be}
> is part of the selbri,
> in particular
> it is part of the second component of the
> tanru, it does not atach a sumti
> to another sumti, the way {pe} does for
> example.

The point of this last remark escapes me
completely.  {be} is used to attach arguments to
selbri when that selbri is not the principal one
of a bridi (here it is the principal one of a
sumti). This does not affect the internal logic
of the structure but refers only to the outer
structure in which it is embedded.  That inner
structure is the same as it would be were the
{lo} replaced by a sumti and the {be} dropped (as
it would be automatically).
 
> > It does instead break into
> > [ralju][bangu be le gligu'e], where now the
> > predicate is broken up as well as the
> arguments
> 
> What do you mean by "as well as the arguments"?
> The predicate
> (i.e. the selbri) is made up of two components,
> yes, a seltau and
> a tertau. Which arguments are broken up there?

As noted, in a simple bridi, the arguments are
regularly broken into two pieces: the ones before
the predicate and the ones after, with the latter
subordinated to the predicate itself.  So there
is an interpretation rule that says that these
two grammatically separated bits are pulled
together into a single logical component (and
component in Lojban, the parser aside). The fact
that the prepredicate argument is absent does not
affect this separation in principle, but here the
oint is just that, just as the arguments are
generally separated, so in this case even the
predicate is broken up.

 
> > (this is all a portion matched by the missing
> > first argument of the selbri).
> 
> Of the full selbri, yes.
> 
> > Now, to be sure,
> > this grouping could make sense, but it is for
> a
> > derivative structure, not the primary one,
> > pred+arg.
> 
> If the pred+arg structure is primary and the
> pred+pred=complex-pred
> is specific to Lojban and absent in FOPL, why
> would FOPL require
> that pred+pred=complex-pred should have
> precedence over
> pred+arg=complex-pred? Is the pred+pred
> structure (absent in FOPL)
> even stronger than a primary one like pred+arg?

That a propsoition divides into predicate +
arguments is fundamental to logic.  A change from
that needs an explanation and, it seems
reasonably, a marker to warn that it is
happening.  As pointed out, the fact that FOPL
does not usually overtly use complex predicates
does not mean that they do not occur.  And when
they do occur, they stick to the
predicate-argument division.  To be sure, within
complex predicates, it is possible to have chunks
which are themselves predicate+argument
structures -- complete propositions -- but these
are always marked.  The present conflict is
between /y/x (x ralju tz z bangu y) le gligu'e, a
simple predicate, and /x(x ralju tz z bangu (be)
le gligu'e) a complex one with a term already in
place within the predicate, which thus needs to
be marked -- and is by conventions about
parentheses and leading lambdas.
In the latter case, the two arguments -- x and
{le gligu'e} are on different levels and are not
to be treated as simple arguments to a single
predicate but rather the first as argument to a
predicate of which the second is already a par. 
Yet (assuming  the corresponding grammatical
analysis of the original phrase is correct) the
two arguments are presented  (assuming the rule
needed for the simplest cases) as being on the
same level.

> > In this structure, the argument to
> > {bangu} really is at a different
> (subordinate)
> > level, down two in Lojban so at leat one in
> logic
> > (given the logical same level is down one
> in
> > Lojban).
> 
> Why is argument absorption by a predicate
> apriori at a lower level
> down than modification of one predicate by
> another. How do you
> figure that? They seem to me to be independent
> notions, and either
> could be defined as having precedence over the
> other.

Look at the structures involved;  In the second,
tthe {le gligu'e} is already in the structure
when the other is added, two levels higher
(parallel with {ralju}, not even with {bangu}. 
In the first the two are both arguments directly
to the predicate, so at the same level, logically
(and one off grammatically).  The relation of
predicate to argument just is primary in logic
and the relation of one predicate to another is
secondary and always depends upon the
argument-predicate relation (see the examples
above).

> > The Lojban that would give this
> > structure without a doubt is {ralju be lo
> bangu
> > be le glicu'e}.  Now, given the
> indefiniteness of
> > sources  for tanru, this might be a source
> for
> > {ralju bangu} but it would be a surprising
> one.
> 
> I think the most common expansion of {ralju
> broda} is going to
> be {ralju (be lo broda be'o) je broda}, i.e.
> the same pattern as
> with {mutce}, {barda}, {cmalu}, etc.

This is a very odd expansion from the point of
view of the patterns developed (not that these
are binding or exhaustive, merely familiar).  As
noted it has some peculiar results, like the
shift in the categorization of x1 between being a
bangu and being a ralju.  This shift suggests
that the two are not equivalent and thus that the
tanru does not come from the longer expression --
or is not to be expanded into it (depending on
how you look at the matter). 
 
> But that has little to do with the general
> question of whether
> tanru composition should or should not have
> precedence over
> argument absorption.

It is attempting to reduce that question down to
one where we know the answer. But, of course, we
already do for logic.  The point is that the
parser seems to have a different rule, which was
the only point I was making.
 
> > We would expect {bangu ralju} on the "lion
> > hunter" model, where  the missing x1 as a
> kind of
> > ralju -- as in the long form, rather than a
> kind
> > of bangu, as in the tanru.
> 
> I sort of gave up on that ideal. I think I am
> now resigned
> to cmalu/barda/mutce/milxe/mutce/traji/ralju
> etc. being
> used as modifiers rather than as main
> components, even
> though their use as main components would give
> a
> simpler expansion.

Well, acknowledging that these take a different
rule is quite enough to make my point.  Notice
that, in general, these can be brought into
normal rules but only at the cost of changing the
parser rule on precedence of predicate argument
and tanru expansion.
 
> > (To be sure, the
> > missing x1 IS a knd of bangu, but that is
> > inferential from the way that [ralju} works,
> not
> > sometyhing said in the form alone.)
> > Since the form we have is not br but rb, we
> are
> > justified (even if ultimately wrong) to take
> it
> > that it has a different source.
> 
> Are you trying to figure out the general
> precedence question
> from the meaning of this particular example?
> There's no guarantee
> that the example is good Lojban.

That is, of course, another point.  It is,
however, grammatical Lojban (even according to
the parser) and so it is relevant to try and
figure out what it might mean.  The two lines of
attack lead to two different but related answers
(indeed, so close that it is hard to figure out
what the practical difference might be in this
particular case). I am notnaturally, trying to
figure out what the precedences are.  I know what
they are for both logic and the parser and that
they are different, which is the point I have
been making all along.  This case merely
illustrates the point (as well as giving a way of
almost legitimating the parser rule in logic --
needing only a marker to be OK, just as the logic
rule is legitimate in the parser with a marker).

> > the obvious one
> > is a "white hunter" tanru "x is a language of
> y
> > and x is a principal one among the languages
> of
> > y", x([rb]y) rather than x(r[by]). This
> > interpretation makes the expression
> explainable
> > under the same rules as are applied with the
> > simple xPy case, whereas the other requires a
> new
> > rule which is nowhere motivated in the
> expression
> > itself, contrary to the general principle
> that
> > deviations from the norm should  be marked
> and
> > the norm unmarked (this latter being violated
> as
> > well, since, in Lojban, the simple case
> requires
> > additional marks).
> 
> I don't follow that.

What part? {ralju bangu} makes good logical sense
with a different tanru structure
(source/expansion), but using that base violates
the parser structure rules again, making the
central point yet once more.

 
> > So, my point that Lojban does a lousy job of
> > representing logic comes down to a couple of
> > possibilities when illustrated by this case:
> 1)
> 
=== message truncated ===
<<> So, my point that Lojban does a lousy job of
> representing logic comes down to a couple of
> possibilities when illustrated by this case: 1)
> the analysis that the grammar gives is what was
> intended, in which case the principle about
> marking is violated

What is the principle about marking?>>

That basic forms are unmarked and deviant ones
marked.

<<>as is the simple rule for
> arguments at the same depth ({le glicu'e}
appears
> to be at the same depth as the missing x1 but
is,
> in fact, at least one level lower)

Appears to whom? Only to someone that assumes
that
tanru formation comes before argument absorption.
If the
tanru had been {bangu be lo gligu'e be'o ralju}
would the
argument absorbed by the seltau still appear at
the same
depth as the missing x1? Or is it only arguments
absorbed
by the tertau that appear to be at the same
depth?>>

The reversed order with the argument attached
inside the tanru is quite enough marking to make
the situation clear (as would be simply inserting
{co} between {ralju} and {bangu} -- though
"simply" is ironic here, a hole restructuring
being involved).

<<> and the
> definition of "truth" has to be suitably
modified
> (despite appearances this is the bear structure
> of a one place predicate, rather than a two
place
> predicate with one place filled).  2) On the
> other hand, if what is meant is the two-place
> relation "is the principal language of," as
> appears from the surface structure,

That's not how the surface structure appears to
me, given
that I know that argument absorption takes
precedence.
Similarly the surface structure of "3 + 2 * 4"
does not
suggest to me that the sum is done first, because
I know
that the convention is that multiplication takes
precedence.>>

My point exactly -- for logic and semantics, the
rule is the opposite.  And, for Lojban (not the
parser) too: a bridi is selbri + sumti and the
bridi is true if the referents of the sumti (in
order) are in the referent of the selbri.  But
under the parser rules, what appear to be the
sumti to a selbri are in fact partly such a sumti
and partly a part of the selbri itself, so the
truth test needs to be redefined somewhat
relative to the surface form.


<<>then the
> grammar's analysis is a total miss -- or, I
> suppose, requires yet another set of rules to
get
> back to what was meant in the first place.  In
> either way, the transparent connection between
> bridi and proposition is even more complicatedx
> than it was in the simple cases.

Tanru expansion cannot be done automatically, but
we
already knew that.

The precedence between argument absorption with
{be} and tanru
formation by juxtaposition of two selbri (absent
in FOPL) is a matter
of convention, not something that can be
determined from FOPL.>>

Well FOPL does not have simple juxtaposition
because it has to be explicit about what the
connection is, so one significant part of getting
from Lojban to logic is working out what that
connection is.  The point is simply that the
parser analysis is often (I would say generally)
misleading about that -- or at least conflicts
with other possibilities that fit more naturally
with logic.  It is not by the way entirely a
matter of convention, sinc every case of
absorption involves a case of application to be
carried out, this giving priority ot application
(which does not generally involve a case of
absorption).