[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

*old response to Jorge



Jorge wrote:
>Well, what is easier:  "The logical connectors appear in four series:
>JA, A, GIhA and GA, here is how you use each series..." or "The logical
>connectors are ja, je, jo, ju, here is how you use them in each
>situation...".
>
>By now I know how to use them all and they don't cause any difficulty,
>but when I first had to learn them I would have much preferred that they
>were not spread into so many selmaho.  Why have to learn four words for
>what is essentially the same concept?

I considered this at one early point in the design.  What became
gi'e/gu'e/zi'e etc. were originally TWO words - ".e" (more basic than
"je"), with a cmavo for scope marker added.

It did not YACC any better (putting SE" on the front and "NAI" on the
end takes a lexer compound, and I was trying to minimize these, and make
the ones needed as simple as possible), and we had CV.V cmavo compounds
as a result, so I went with building the scope in as a morphological
element within the connective.  You could still analyze gi'e as
"gi'"+"e", if you want, and this isn't bad pedagogically since it casts
a direct spotlight on logical scope of connectives, which is the whole
point for the multitude of words (blurring of usage of "je" necessarily
blurs the obviousness of the logical scope, which is in the final
analysis why I see any combining "ambiguous").

I did not consider "je" as the basic form - it is after all "j-"+"A", so
did not consider compounds like "jije/guje/zije" which might have worked
(but still would have been hard to YACC).  This would have eliminated
the phonological problem of the mandatory intervocalic pause, but not
the other problems.  Ultimately I went with the status quo - separate
words, because JCB already had the equivalents of A and JA, and adding
new series thus seemed to be more historically Loglandic (and hence more
likely to survive the ever-hoped-for reconciliation negotiations of the
time) than having cmavo compounded to indicate scope.

I think this is one point that has fallen through the cracks on the
debates about design:  One major reason for doing things "like JCB did"
is that during most of the Lojban major design phase, our decisions were
guided by the expectation that the very act of showing that we could
design Lojban would force JCB into compromise, and our strongest
negotiation position for reconciliation with him would occur if we could
point to similarities with his design.  Then with the trademark lawsuit,
it became even more important to continually show that Lojban "was
Loglan".  We never had the sense that we were trying for anything new
and independent, and did not have people in the community with such a
concept until well after the major design was pretty firm.

>lojbab:
>> JCB had only one non-logical connector, which he used in a trivial way in
>> tanru and nowhere else.  We have ce and ce'o at minimum that have seen wide
>> and necesary use,
>
>Wide and necessary?  I never use them, since I don't use sets in normal
>talk.  I don't see others using them very often either.

There are gismu place structures that require sets in some sumti. ce'o
is the only way to imply any ordering that is not time- (or possibly
space-) based.  I imagine that .ice'o will be used for ordered lists of
sentence-level constructs - something that isn't used too much in the
type of communications we thus far have written in the language.

>> in addition to joi - which encompasses JCB's connective
>> but may be being overstretched in current usage.
>
>{joi} is extremely useful.  In fact, it should be used much more often
>where people tend to use {.e}.  The collective plural is more basic than
>the distributive one.  (And I'm happy to have found a statement in
>McCawley's book agreeing with that.)

That may be so, but it is a non-logical construct.  To accept McCawley's
statement at face value in effect rejects the Loglan hypothesis from the
start.  Indeed, since languages other than Loglan/Lojban exist without a
singular/plural distinction, I find anything that states that a plural
is "basic" much less a particular kind of plural to be somewhat biased.

>> Actually, though, since you ask, I suspect that the LE for jo'u would have
>> many similarities to lo'e/le'e which focus on the commonalities in a
>> disparate set, which is what I THINK jo'u does.
>
>I don't really see it.  Are you saying that I could use {le'e re broda}
>instead of {le vi broda ku jo'u le va broda}?

Not sure, as I said.  You still have some intensional aspects in the two
"le"s that seem lost.  le'e vijo'uva broda would work, I think.

What I am more sure that I am saying is that if we had a set of LE
operators le'o'u, lei'o'u, le'i'o'u, le'e'o'u corresponding to le, lei,
le'i, le'e but having the semantics of jo'u, then le'o'u broda would be
something that had the common properties of broda (i.e. lo'e, not le'e).

>> > You don't have
>> >to use {je} instead of {.e} if you don't want to.
>>
>> Yes, but you are asking for errors and noisy-environment ambiguity, which we
>> can tolerate minimally in joi connection, but not in logical connection.
>
>Have you done any research to support that statement?  Nothing would
>stop you from using {.e} in noisy environments, but I don't see why
>errors with {joi} are any more tolerable than errors with {je}.

Because 1) joi is a non-logical connective, so Lojban practically
doesn't CARE if they are logically flawed, and 2) claims involving joi
tend to be rather weak mass claims truth-functionally even if you TRY to
force a logical usage onto them.  They are, in essence an "necessary
evil", an add-on to the langauge to support "lesser" features that must
be tolerated in a natlang but are really beneath the dignity of a
logical language.  You usually use them when you are tolerating
sloppiness in the first place, so a logical error is probably less
offensive there as well.

>> Pragmatics can usually clarify an error in JOI connection; I doubt that
>> it will be as robust for logical connection.
>
>Why?  I don't see any possible reason to support that.  Why would one be
>easier to clarify than the other?  The labels "logical" and
>"non-logical" are purely Lojbanic conventions.  There is nothing more
>logical about one type of connection than the other.  It's more a matter
>of distributiveness than anything else.

NOT having read McCawley, alas (still on special order), at least I can
say that the stereotype of predicate logic (if not the reality) is that
it involves the logical connectives. pc has seemed to say that plurals
done the natlangway , or even the Lojban way, are logically sloppy - the
"logical" way to express plurals is very long-winded and DOES involve
logical connections and operators.

Put another way - there is a predicate calculus symbol for the logical
connectives.  There is, AFAIK, NO such symbology for any of the JOI
connectives.  Thus the convention is NOT Lojban's, but that of predicate
logic/predicate calculus.

>> It will become too easy to make
>> mistakes with logical connectives (which are a critical part of the
>> logical-language core) that are grammatical but don't mean what you want.
>
>Examples?  Lojbab said something similar but I don't see how that would
>be.  Why would such mistakes (if they indeed are so easy to make, I
>don't really see how) be more critical with {je} than with {joi}?
>
>> Lojban is altogether too rich in such things already.
>
>Lojban has an excess of stuff in some parts, I agree, but I don't see
>the problem here.  The change would simplify the system, by virtually
>reducing the number of selmahos, not complicate it.

The argument is one of linguistic redundancy, of which Lojban has
precious little.  We first note that virtually EVERY CV and CVV syllable
(or disyllable) is a valid cmavo, and something over 80% of the rafsi
(which overlap the CVV syllables enough to make the tosmabru test
necessary) have valid meanings.  As a result, virtually every string of
CVs CVVs and smatterings of CVCs and CCVs is lexable, even if it might
not be grammatical.

Furthermore, fu'ivla space is defined negatively, such that basically
anything that doesn't lex as a cmavo or gismu or lujvo is a valid
fu'ivla, perhaps nonce, of uncertain meaning.

It thus turns out that the relatively unpacked gismu space is the
biggest hedge we have against insufficient redundancy at the word level,
and this only works if you allow no error correction for the occasional
1-letter errors that we see in actual usage.

As a result, Lojban's redundancy is almost SOLELY provided at the
grammatical level.  At this level, brivla and certain other words (like
na'e and ke and co and nu) can occur with little grammatical restriction
on order in a selbri (or description).  And will of course choke at the
concept of "ta na'e nu no'e du'u to'e nu broda ke na'e du'u nu brode",
but it is perfectly grammatical (and And insists that anything
grammatical according to theory is by definition not nonsense even when
no native speaker can figure out what it means %^).

Since all of the redundancy occurs at the grammatical level, we then
need some sense of the redundancy level of the grammar.  We don;t have
any direct one, but a good sense of the grammatical redundancy can be
seen in the elidable terminators.  To the extent that terminators are
NOT required, the grammar is fairly non-redundant, since the need for an
elidable terminator to actually be used in a sentence is tied to the
existence of multiple selma'o that can occur immediately following that
terminator.

As such it is easily arguable that any grammar change that forces the
elision of a terminator to occur less frequenty is a direct loss in
redundancy.

Now many of the selma'o in the language have immense grammatical
ambiguity on their own?  ".i" at the beginning of a sentence can
probably be followed by at least half of the selma'o in the language
without necessarily causing a grammar error.  At those places,
redundancy is at a minimum.  Other places include anywhere a selbri or a
sumti/term can occur.  Guess what - after je, joi, and .e are just such
things.  But right now, the je and the .e follow-ons are largely
complementary since the grammar has to be able to resolve sumti and
selbri VERY earlier accordingg to YACC LALR rules.  But the proposal
would change that, and after JA, as with after JOI, you would have an
enormous range of grammatical possibilities.  The proper use of
elidables MIGHT help this, but I distrust people using elidables ideally
- it has proven to be one of the more difficult features of the language
to teach.

>More or less related:  does {jimpe} mean being in a state of
>understanding something, or does it mean to come to understand
>something?  "Understand" can have both meanings in English.  Does {mi
>ba'o jimpe} mean that I no longer understand or that I've already
>understood something?

The former.  jimpybi'o would be "come to understand"

za'o jimpe is an interesting concept %^)

lojbab