Chief logician?
Randall Holmes
respect to the Loglan project because of my relation to TLI. I
disclaim any such role. My job over here is mostly critical; since
the language is "ready for use", I can't do major engineering
modifications on it
Logical Language Group
RH>respect to the Loglan project because of my relation to TLI. I
RH>disclaim any such role. My job over here is mostly critical; since
RH>the language is "ready for use", I can't do major engineering
RH>modifications on it
I wouldn't disagree with YOU %^)
But
1) JCB says that Loglan will never be done and reserves the right to make
changes by decree indefinitely into the future. See Loglan 1, and the
charter for his Academy - there are NO restrictions on what changes are
acceptable, and JCB has explicitly DENIED that the community ('the masses' in
his lingo) are relevant to such decisions, which are to be made on purely
ummmm 'academic' grounds by his 'judicial' Academy.
2) A statement that the language is "ready for use" is not true until it
has been put to the test of significant usage. Lojban has had close to
the level of usage that I can start feeling comfortable with such a claim,
but it is the breadth of our user base as much as the amount that any
individual has written that convinces me more than anything else. TLI
Loglan has proven capable of expressing a limited set of things that
JCB and a few other people have tried to say with it, but as far as I know,
none of them are non-native English speakers, and the numbers are too small
to really even claim that they cross-section the English speaking world
very well. (And JCB's linguistics research doesn't give much confidence
either.)
AS an example - in the purely logical arena - I suspect that JCB has never
done the anlysis of negation needed. He has handled logical negation,
of predicates, but people pointed out to us early on that there were a lot
more variations on negation than we had examined. So pc went through Horn's
"Natural History of Negation" which conveniently had just been published, and
we made some significant enhancements to the language as a result. I have
heard nothing abouyt similar changes to the TLI language. All human languages
seem to have some features, but it is not clear that TLI Loglan has them.
3)Determining the semantics of Loglan (all versions) is a somewhat open-ended
job. Most of our debates in the Lojban community have been about semantics,
and not the 'engineering' level of the language. I would presume that you,
Randlal, have significant impact on the semantic interpretation of statements
in TLI Loglan, if such statements fall within the domain of analysis of logic.
lojbab
Jorge Llambias
> Does anyone in the
> Lojban community realize that logical connectives applied to arguments
> produce problems of scope (usually handled implicitly in NL's)
> precisely analogous to those connected with quantification?
Logical connectives are never applied to arguments in Lojban. (In the sense
that all forms of logical connection are contractions of full bridi
connections.)
And yes, the debate on "any" was sparkled by a logical connection question:
mi nitcu le tanxe a le dakli
I need (the box OR the bag)
means:
(I need the box) OR (I need the bag)
It does not mean what we usually mean in English by "I need either of the
box or the bag".
> Consider
>
> John and James love Mary or Sally
>
> versus
>
> Mary or Sally is loved by John and James
>
> In the second sentence, but not in the first, it is clear that John
> and James love the same unspecified element of {Mary, Sally}; in the
> first sentence, they may love different elements of the set.
If you have two logical connectives in a Lojban sentence, I think the
first one binds tighter, so
la djan e la djeimyz prami la meris a la salis
expands to:
la djan e la djeimyz prami la meris
ija la djan e la djeimyz prami la salis
(John loves Mary AND James loves Mary)
OR (John loves Sally AND James loves Sally)
while:
la meris a la salis se prami la djan e la djeimyz
goes to:
la meris a la salis se prami la djan
ije la meris a la salis se prami la djeimyz
(Mary is loved by John OR Sally is loved by John)
AND (Mary is loved by James OR Sally is loved by James)
This is in reverse of the meaning you give for the English sentences,
but there is no ambiguity.
Jorge
Logical Language Group
RH>an academy will become about as relevant as those of the Academy which
RH>imagines that it legislates for the French language, regardless of
RH>what any individual thinks.
Agreed. Buit JCB doesn't agree, and there's the rub in a society where he
controls all the strings. In TLI society, if JCB and his academy don't
like how you use the language, his argument is that their copyrights and
(former) trademarks give them the right to control the usages made of the
language. Indeed, *IF* TLI had kept the tardemark, they would not only
have had the right, they would have had the RESPONSIBILITY to police all
usages that were labelled 'Loglan'.
But we need not discuss politics if you do not wish to %^)
RH>A project which both languages should consider is the mechanization of
RH>not only the grammar of the language but of the allowed logical
RH>transformations; this would make it possible for interaction with
RH>machines to enforce the logical usages.
I'm not competent to comment on the desireability or usefulness of this.
Others in the community can do better.
RH>It is far easier to
RH>learn to utter Loglan sentences which parse correctly than to learn
RH>the logical (and philosophical!) background knowledge needed to use
RH>the language(s) correctly.
Hey - you've seen the level of argument in this current discussion. We've
got people who seem to know how to dicuss these types of problems
intelligently, and for the most part independent of their knowledgfe of English.
But I think that MOST Lojban usage is relatively independent of the logical
problems. People just do not make statements that involve quantifiactional
variables for the most part. There is a nearly complete lack of the kind of
universal statements made in English colloquial conversation, for example.
People use "le" with its intensional constraints, and that pretty much
eliminates this whole question of existence and sets and their membership
for those usages where "le" is acceptable.
I won;t say we are perfect, but we are doing pretty darn good.
An example - while we are now arguing about "need" (which I think is the
real problem in "I need a box", not the "a"), we HAVE tackled "seek", and
you no longer seek objects that may exist - you seek property abstractions.
We have also dealt with raising of abstractions, a natlang phenomenon that is
endemic and invisible, and a serious logical and philosophical problem in
TLI Loglan. WE have the solution, and people are using it with far more
sophistication and 'anturalness' than one might have thought.
RH>My "feel" for both languages is that they are too
RH>similar to the native languages of the experimenters; see above. If I
RH>were designing a language from scratch, I would have adopted VSO or
RH>even OSV word order (Polish or reverse Polish notation :-) ), for
RH>example.
Well, what is the native language of our 'experimenters'? Yes, I am hopelessly
English-bound. But the people who have been engaging you in the current
discussion include Jorge Llambias, who is Argentinian, and Veijo Vilva who
is
Finnish. Nick Nicholas in Australia is bilingual-native Greek and English
and is the most fluent user of Lojban these days. He is also skilled in
Klingon, which is as un-English as they come, and leads their philosophical
discussions as well as ours. We are also starting to get input from Chines Lojbanists.
RH>I don't think that the scientific or non-scientific nature
RH>of JCB's method for contructing primitives, for example, is at all
RH>relevant to the usability of the language.
Agreed. This is a 'marketing gimmick' in some ways. But you need some
non-random method of getting words that are different from native language
words, so that you can fight automatic transfer of semantics. We seem to be
doing fairly well at that battle, though it takes eternal vigilance.
It does offer some possibliity of linguistics research into factors that
affect learnability of vocabularies, but this is really incidental to the
actual usability of the language for many of its purposes.
RH>Negation is fine in Loglan; except that I'm not sure I would have allowed
RH>negation of arguments; this has no analogue in symbolic logic usage
RH>and can lead to very misleading transformations.
Please discuss handling of scalar vs. contradictory negation in TLI Loglan
sometime %^) Then try handling metalinguistic negation (for thatmatter try doing
much of anything metalinguistic in TLI Loglan).
RH>Lojban community realize that logical connectives applied to arguments
RH>produce problems of scope (usually handled implicitly in NL's)
RH>precisely analogous to those connected with quantification?
RH>
RH>Consider
RH>
RH>John and James love Mary or Sally
RH>
RH>versus
RH>
RH>Mary or Sally is loved by John and James
RH>
RH>In the second sentence, but not in the first, it is clear that John
RH>and James love the same unspecified element of {Mary, Sally}; in the
RH>first sentence, they may love different elements of the set.
These sentences in LOjban will expand into a series of sentences that are
logically connected. Negation may require use of DeMorgan's Law. The
Lojban negations should cause few problems.
We do not have logical negation of arguments though - we have contrary
negation (this is my off-the cuff recollection). You can talk about
going to 'other-than Rome'. If my recollection is incorrect, though,
we still have dealt with the questions and come up with answers which are
not the natlang answers.
lojbab
Randall Holmes
much of an invitation to political argument. Betrayed by my mailer...
If Loglan or Lojban becomes a language really used, the decisions of
an academy will become about as relevant as those of the Academy which
imagines that it legislates for the French language, regardless of
what any individual thinks. There is a proper function for an
advisory body in the area of correct logical usage, however. My
suspicion is that the logical features of the language (either
language) would collapse under the pressure of widespread usage, and
will become closely analogous to "bad" NL usages in hard cases; the
_option_ of logical clarity will remain. Machine parseability (and
maybe the logical usages to some extent) could continue to be enforced
if a major part of the speech community consisted of computer
programs.
A project which both languages should consider is the mechanization of
not only the grammar of the language but of the allowed logical
transformations; this would make it possible for interaction with
machines to enforce the logical usages. In the limit, the
construction of a theorem prover with (possibly subset) Loglan/Lojban
as the input language should be considered. I have indicated to TLI
that work of mine (not motivated by Loglan/Lojban) may make it
relatively easy to construct such a program in a few years; it is
probably already possible with "off-the-shelf" techinology using
existing theorem prover systems.
Note that I put "ready for use" in quotes (I think I did). But, in
fact, the language (the despised TLI Loglan, that is) is ready for
use. I have done enough translations into it to be fairly certain of
this. I am not certain that it could be spoken correctly by anyone
without something approaching my own peculiar qualifications, and I
think that the same is probably true of Lojban; the pressure of the
usages of the natural languages is too powerful. It is far easier to
learn to utter Loglan sentences which parse correctly than to learn
the logical (and philosophical!) background knowledge needed to use
the language(s) correctly.
I won't express an opinion of JCB's linguistics research; I'm not
competent to do so. My "feel" for both languages is that they are too
similar to the native languages of the experimenters; see above. If I
were designing a language from scratch, I would have adopted VSO or
even OSV word order (Polish or reverse Polish notation :-) ), for
example. I don't think that the scientific or non-scientific nature
of JCB's method for contructing primitives, for example, is at all
relevant to the usability of the language.
Negation is fine in Loglan; except that I'm not sure I would have allowed
negation of arguments; this has no analogue in symbolic logic usage
and can lead to very misleading transformations. Does anyone in the
Lojban community realize that logical connectives applied to arguments
produce problems of scope (usually handled implicitly in NL's)
precisely analogous to those connected with quantification?
Consider
John and James love Mary or Sally
versus
Mary or Sally is loved by John and James
In the second sentence, but not in the first, it is clear that John
and James love the same unspecified element of {Mary, Sally}; in the
first sentence, they may love different elements of the set.
--Randall Holmes
pet logician, TLI :-)
Randall Holmes
intelligently. So do we. There is a difference between doing this
explicitly and doing it implicitly and more or less automatically, as
one must in speaking a language. You address the question of how this
goes on below, I see; I'm very interested in how the use of the
machinery works out in practice, but it seems that the answer is "No
one uses it"!
On people not making statements that involve quantificational
variables -- that's a practical measure of how well people have
mastered the machinery, I suppose :-) Has anyone tried to speak or
write about mathematics in Lojban at a level where this becomes
unavoidable?
Expand your comment about "the raising of abstractions"; I think I
know what you are talking about (and that it is important) but I am
not sure I understand you.
The only genuinely alien languages you list among native languages of
Lojbanists are Chinese and Finnish.
On your comments about negation, I have problems with terminology, but
as far as I understand you I probably disagree flatly. There is one
concept of negation (the propositional connective) and then there are
various other notions which NL's confuse with negation; I would hate
to think that you are importing NL confusions (more likely you are
defining these other notions precisely and using them correctly and
the only confusion is that you call them "negation" :-) ). Explain by
example what you mean by "metalinguistic" negation.
Of course I understand how logical connectives applied to arguments
are eliminated! The difficulty arises in expanding sentences when
there is more than one such "argument" in it, and an answer I received
seems to indicate that you have an official solution to this (good! --
so far as I know, TLI Loglan does not) but that it goes contrary to
the natural analogy with implicit quantification (not so good --
explained fully in another post)
--Randall Holmes
Jorge Llambias
is responding here]
> On your comments about negation, I have problems with terminology, but
> as far as I understand you I probably disagree flatly. There is one
> concept of negation (the propositional connective) and then there are
> various other notions which NL's confuse with negation;
I agree.
> I would hate
> to think that you are importing NL confusions (more likely you are
> defining these other notions precisely and using them correctly and
> the only confusion is that you call them "negation" :-) ).
In fact, I think that's exactly what's going on.
For example, does Loglan have {na'e} = non-/other than ?
We can easily distinguish
ta na blanu tanxe
It is false that: that is a blue box.
ta na'e blanu tanxe
That is a non-blue box
and things like that.
I don't like calling {na'e} negation, and even less calling {to'e} negation
(to'e=opposite). But they are very useful.
> Explain by
> example what you mean by "metalinguistic" negation.
It's the answer to "Have you stopped beating your wife?"
Strictly logically, I think that {na} suffices, but that leaves the
wrong impression in some people.
> Of course I understand how logical connectives applied to arguments
> are eliminated! The difficulty arises in expanding sentences when
> there is more than one such "argument" in it, and an answer I received
> seems to indicate that you have an official solution to this (good! --
> so far as I know, TLI Loglan does not) but that it goes contrary to
> the natural analogy with implicit quantification (not so good --
> explained fully in another post)
I'm not sure if there really was an official solution, since that
particular case is not explicitly mentioned in the connectives paper.
I agree that the opposite order to the one I suggested may be better.
Jorge
Randall Holmes
you were referring to; it is indeed quite important. TLI Loglan has
the problem you have described with implicit raising of arguments in
predicates; I have already noticed it. We may eventually need to do such
a cleanup; I'm happy to hear that it proved to be possible (or was it a
success :-) )
I also support using real terminaology for negation; for instance, having
remarked
that I felt that negation properly was an operation on propositions,
I did go ahead and speak of negation and other logical connectives as
applied to arguments... [A [B But the other forms of negation you describe
(other than the metalinguistic, which I suppose really is a logical
negation) could use new terms. Sometimes an existing usage is actually bad
and needs to be reformed; isn't the whole Loglan enterprise an extremely
radical move in that direction :-)
--Randall Holmes
P.S. Could we stop replying to messages with this heading? It makes
me itch, as if I had some unfulfilled responsiblity somewhere...
Logical Language Group
far, for the most part, people haven't had much trouble.
For one thing, we have more explicitly defined negation as a predicate
negation that has full sentence scope, whereas natlang negation tends to
only have scope over 'the rest of the sentence' following the negation.
We define our negation as being immediately exportable to the prenex by
definition without change inquantifiers. In my own usage, therefore, when
I rarely have to resort to quantified variables, I just leave the negation
in the prenex and do not use the 'short forms' of natlangs where the
variables are not explicitly quantified. I find this form easier to
understand, almost error-free in manipulation, and the circumstances where
I have been required to use it rare enough that Zipfean shortening doesn't
seem to enter in.
(We do have a natlang style negation form, but it isn't much used except in
translation.)
The easiest way to exemplify our variuos forms of negation is to refer you
to John Cowan's negation paper on ftp.cs.yale.edu pub/lojban/draft/refgrammar
At last report there was an error in one section (a DeMorgan expansion)
that might not have been caught, but it has lots of deatails of the various
forms of negation.
If I recall the discussion, the two more important kinds of negation are
defined by reference to a classic (I think Aristotle) 'box' diagram, where
contradictory negation is expressed by moving diagonally across the box, and
contrary negation by moving horizontally (ort is it vertically and horizontally
or ...).
I am not walking to the store
could mean
1) simply that it is false that I am walking to the store
2) that I am driving or running to the store
1) is contradictory negation
2) is contrary negation
Metalinguistic negation deals with things like
Have you stopped beating you wife?
when you haven't started - it says that there is some hidden assumption that
is incorrect that makes the sentence not only false but meaningless from a
truth-functional point of view.
Contrary/scalar negation works very well on arguments:
I don't have 3 children (I have 2 or 4) is expressed in Lojban better
by nbenegating the argumetn "3 children".
Metsalinguistic negation crops up in a variety of ways. Grammatically in
Lojban it is an attitudinal (actually a discursive - but same grammar).
Contrary negation can be applied to arguments, can be applied to predicate
words within a metaphor, or can be applied to an entire metaphor or
predicate 9and a few other things). Contradictory negation only applies to
sentences, and all forms of contradictory negation of a sentence negate the
entire sentence (Modulo the special natlang scope form which is used to
exempt quantified variables in a 'natural way')
====
Abstract raising has reared its uglu head in so many ways in Loglan/Lojban
that I hesitate to try to cover all of them. One is underlying the current
discussion on nitcu/need. Do you need an object, or do you need a state/event
involving that object?
Causality, is it 'guns' that kill (dead-cause) people, 'people' that kill
people, or 'people shooting guns' that kill people. The first two are
probably raisings of the third. Most Lojban causality is now expressed
as 'event causes event' rather than 'object causes event' or 'object causes
object' or 'event causes object'. But we have a form, now commonly used that
recognizes that we aren't always sure exactly what the event is, or how to
express it (or simply that it doesn't matter), so we can explicitly raise
the argument from abstract to object with a marked form.
e.g.
my work is done
le nu mi gunka cu mulno (lojban)
lepo mi turka ga kapli (TLI)
and not
*lemi gunka cu mulno
?lemi turka ga kapli
but Lojban allows
tu'a mi mulno
raise I am-done
"I'm done."
which if the L1 place styruture is correct
"mi kapli" should never be said, unless perhaps at one's point of death
if you want to view yourself as an event.
We had to go through all of the place structures and decide which of them
took events as a norm, while which took objects (there are circumstances
where cossover is permitted and works without marking though), and in some
cases mosdify place structures to eliminate raising. TLI Loglan words
that have hidden raising are most evident in predicates where there is
an abstraction in one argument, and a concrete (often an agent) in another
argument, such that the concrete is ALWAYS found as a place in the
abstraction. This is not always 'bad', but we have found it worthwhile to
know exactly where it occurs, because logical errors seem to explode out of
such constructions. (For one thing, quantified variables inside an
abstraction do not necessarily export to the sentence level. But if the
two levels are muddled, you can get errors in scope.)
====
Oh, I failed to address one point under 'negation' - termionology. We much
prefer to use 'real' definitions of terms rather than jargon definitiuons.
Only to a logician does 'negation' apply only to propositional forms. English
words have English meanings, and most people think of 'negation' as being
tied to the meaning of "not" in a wide variety of usages.
You can control discussions by insisting on narrow meanings for words, but
then you find that the results of the discussion do not have wide applicability.
WE chose instead to bite the bullet and tackle the fuzzier issues of negation
along with the purely propositional ones, because then people know how to
handle the concept "not" even when it is not purely a propositional usage.
lojbab
Jorge Llambias
> There is a good reason for the precedence to go the other way,
> unless you also reverse the usual convention for implicit quantification:
>
> the point is that P[John and James] means roughly the same thing as
> "For all x in {John,James}, P[x]", and, similarly, P[Mary and Sally]
^^^
or, I think
> means roughly the same thing as "For some x in {Mary,Sally}, P[x]";
> where two of these connected arguments appear in a sentence, one has
> essentially the same problem one has with the usual form of implicit
> quantification as in
>
> Someone loves everyone
>
> versus
>
> Everyone is loved by someone
Yes, I see your point. In fact, the paper on connectives doesn't really
mention that case. It says that
broda ije brode ija brodi
is grouped from left to right:
(broda ije brode) ija brodi
and from there I generalized to assume that the first connective is
bound tighter.
Maybe it should be the other way around when the connectives are in
different terms.
There is still going to be counterintuitive cases, though:
da prami la djan e la djeimyz
Someone loves John and James
means the same as:
la djan e la djeimyz se prami da
John and James are loved by someone
in both cases, the quantification is: For some da; for all x in {John; James}.
> Of course, I know that the underlying "expanded" form of the sentence
> does not involve application of logical connectives to arguments;
> I'm a logician, remember?
Of course :)
But you did mention something about negation of arguments being allowed
in Loglan...
Jorge