response on old posting - lujvo-making TECH/PHILOSOPHY (long)

[Logical Language Group <lojbab@ACCESS.DIGEX.NET>]

I may have finally (re?-)crystallized in my mind what it is I don't like
about the evolution of the lujvo-place-structure effort that Nick and
others have worked on.  I support the work that is being done - it is
giving useful analytical techniques that will in turn lead to more
natural, quick, expansion of the lujvo vocabulary in the language.  On
the other hand, I see a danger of overanalysis and overprescription
which will straightjacket people learning and using the language,
especially those who have not mastered the techniques being described,
and/or who don't know all the place structures.

Cowan weighed in on zi'o and place deletion in lujvo over a month ago as
follows, responding to an anonymous source:
>Lojbab replied to his correspondent thus:
>>There is every expectation that some places will fall off from disuse,
>>but this is what language evolution is about.  If the place falls off,
>>though, under the definitions of predicate logic, it is a different
>>predicate.  It has different truth conditions, etc.  All of predicate
>>logic is based on the foundation that two predicates are not identical
>>if they have different numbers of or sets of arguments.  I can't change
>>this by fiat even if I wanted to, and Lojban has to be concsistent with
>>predicate logic to have the claim of being a "logical language".
>
>This is true only if the place really goes extinct as opposed to being
>almost never filled (which is the status of the x2 place of "gerku"
>anyway:  how often do we think of "gerku" as a 2-place relationship
>between an individual dog and its species/breed?).  I think it is
>entirely safe to let such places maintain a shadowy existence, perhaps
>being filled only in exceptional circumstances.  Once we officially
>delete them, there is no going back.
>
>In sum, I am becoming more and more convinced that the whole idea of
>"deleting a place because it's irrelevant to the [English or other NL]
>concept" is the most stinking of red-herrings.  We should remove only
>places that clearly overlap other places or that are implicitly filled
>by other lujvo members taken as events (in belenu-lujvo, e.g.).

This argument has the problem backwards.  Indeed for examples that we
use NOW, we are talking of deleting places to match a natlang concept.
But in pure Lojban thought, without recourse to translation, we want to
be able to freely create relations between concepts.  So when I want to
talk about a motion predicate CONCEPT that is in my mind, one that links
a thing moving, a destination, and let us say a means of locomotion, and
I wish to make no other implications of relation than these three
things, how do I express it.  The concept is similar to that of
'arrival' and I would use that as the keyword, just as I use 'leave' for
'cliva', but these are Lojban predicates, and not translations of the
English words, or necessarily mappings to English concepts (there are
'leavings' that are not 'nu cliva'; perhaps there are 'nu cliva' that
are not 'leavings', though I'm not sure I could come up with an
example).

The problem that resulted in zi'o is how to build a predicate that
relates exactly the sumti you want, making no implications you don't
want.  This seems vital in logical discourse.  It is the opposite side
of the coin from the insistence that a klama always has the five defined
places.

We then turn from this problem to the corresponding one for
lujvo-making.  Thinking in Lojban, I conceive of a concept.  I try to
make a lujvo for it.  But the 'fat lujvo' advocates that would insist on
all places being present, mean that the choice of almost any gismu to
add into my (presumably lean) concept, will add one or more extraneous
places to the representation of my concept.

Now I think it GOOD that Lojban force me to consider whether, for
example, a dog house maker has in mind a particular breed of dog.

But if he does not, and in fact the house is intended for 'loi gerku' of
whatever species happens to end up using it - or maybe no dog will
actually end up using it, and the house is just there to LOOK like a dog
might use it (with a nice BEWARE OF DOG sign posted), then the fact that
I can theoretically fill the place with 'loi se gerku' is really rather
pointless (indeed, this is one obvious criterion where Nick might
specifically identify possible lujvo place deletion - when the value of
the place is 'loi [the corresponding gismu place]' without any
restriction.


Another quote, this time from Jorge at about the same time
>la lojbab cusku di'e
>> If in thinking a certain Lojban thought you do not perceive
>> a relevant relation with a sumti of semantics x-sub-n, you should be able
>> to omit that place from the place structure of the lujvo you are coining
>> to express that relationship AT THE TIME OF COINING IT.
>
>But how can you not perceive it if it's part of the gismu's meaning?
>Are gismu a relationship between all of it's places, or just a
>collection of relationships between the places taken as single places,
>as pairs, as triplets, etc.
>
>>If your concept truly doesn't include something in the relationship, you
>>should be able to leave it out.
>
>Or question yourself how come your concept doesn't include it, but the
>underlying gismu does.  Is it the appropriate gismu?  Maybe it's a fat
>gismu, because a gismu without that place would make a lot of sense?

I want lujvo-makers to _conceive_ first, then make lujvo to match their
conceptions.  I don't think they should be constrained by our ad hoc,
often arbitrary, and probably cultrually-biased decisions about gismu
place structures.  I also don't want them to be forced to either go to a
natlang and argue from the principles of that natlang what the Lojban
place structure should be, or be forced to leave all places present just
because it make lujvo-interpretation easier (which I doubt would be the
case anyway, if too many places hang around).

When you are making a lujvo, it should be irrelevant whether the gismu
you are making it from is 'fat'.  The lujvo is NOT the gismu; it is a
different, albeit (hopefully) related word.

>>  This because not only do we not have a value
>> in mind for the place in question, but if challenged, we might indeed agree
>> that the relationship we DID have in mind might NOT require some value for
>> the omitted place.
>
>And all I'm asking is what happened to that place.  At which point of
>the lujvo making was it lost.  How come the gismu is present if that
>place is meaningless?

The place wasn't "lost" to me because that isn't how >I< make lujvo.
The gismu suggests a meaning - that is the nature of metaphor.  There is
no particular likelihood that some other gismu will have the same place
structure but one less place, and you shouldn;t be compelled to try to
find one - especially when you are making a lujvo on a pattern-transfer
basis, as I suspect many will be.

A similar argument I've often seen (no quote handy), especially in
discussions between Nick and Jorge, is that it is possibly biased to
delete some places but not others, and that this implies that 'the wrong
gismu is being used'.  But the alternative would be a requirement that
there be gismu for all possible combinations of sumti with and without
other sumti present, to make it possible to form whatever place
structure you want to express.

This is going too far the opposite end of the scale from JCB, who has ad
hoc place structures for compounds that bear no resemblence to their
components at times, since he is using malrarna (usually malglico)
metaphors to express his lujvo.  In designing Lojban, I have made a
strong case that lujvo be fairly literal, and that you shouldn't
willy-nilly introduce places that are not related to any of the
lujvo-components.

Even there, however, >I< would want to occasionally accept unrelated
places provided that there is a meaningful 'term-deletion' from a
specific longer-lujvo taking place to make the resulting lujvo shorter.

But the more (literal-lujvo)ists want totally predictable place
structures and an absence of the possible natlang bias anything else
might permit, or even encourage.

I think such predictability would suffocate the langauge though.  I want
to be able to make a short word that reflects a reasonably-simple and
frequently-expressed thought/concept/relationship.  This means I want in
lujvo MAKING to be able to ignore the analysis that Nick wants, and make
the lujvo I want to, on an ad hoc basis.  The only analysis I feel I
need to make is a) remember if there is another concept that I know that
already uses that lujvo, but perhaps with a different conceptual intent,
and b) try to imagine from the source tanru for the lujvo being made,
whether that tanru might suggest some other, incompatible, concept THAT
WILL LIKELY SEE USAGE.

These two analyses are basically looking backwards, and looking
forwards, with regard to usage of the lujvo in question.  We are trying
to be considerate of other speakers of the language who might use the
word differently than we are wanting to right now.  But the essential
point is to get my idea across NOW, and too many extraneous places,
and/or too many extraneous terms makes it more difficult for such
spontaneous usage of the langauge to take place.

Yes, your remembrance of what has come before will be incomplete and/or
inaccurate, and likely biased by your language background.  Similarly,
your extrapolation forward will be at least potentially biased.  But
only a totally algorithmic language will eliminate such biases
completely, and we already have ruled that out as a plausible goal.
Lojban is NOT algorithmic, nor even close, in how large numbers of its
features will work in actual usage.  I don't even favor trying to make
the language algorithmic, which is what I label jimc's effort as being,
because it gives the language an unntural image and 'feel'.  And that is
already a problem for Lojban as a 'logical language' in the set of
'planned languages' each of which terms implies some degree of unnatural
regularity that turns off some people who might really be interested if
they weren't limited by the image.

On the other hand, if we have taught good lujvo-making HABITS, then a
lujvo will be quickly made, which will be close enough to algorithmic
that a listener should be able to analyze the lujvo using algorithmic
techniques, apply some common sense, and confirm what context says about
the place structure.  Such analysis might then suggest a better word for
the concept, as often happens when people post text here on Lojban List.

What little skill I have manged to acquire with Lojban, comes from the
fact that I CAN and HAVE made up lujvo on the fly when I speak and
write, without recourse to algorithms, and I do so at a speed that makes
much of my expression have some aura of spontaneity, though even I
cannot devise tanru and turn them into lujvo at fluent speeds.  I am out
of practice now, especially not having studied rafsi since the remaking,
but I did do this at close to acceptable speeds for conversation.
Acceptable in this context seems like it should be comparable to the
time a natlang speaker spends groping for a word for a concept,
discovering that he can't think of one, and then using whatever
acceptable rules he knows like Latinate prefixation and verbing nouns
(each of which includes a self-defining-by-example-coinage).  I think
around 10 seconds is a plausible limit, and the speaker must be without
recourse to dictionaries and without requiring complete memorization of
place structures should deal with most such cases.

If Lojban is ever to become a real spoken language, people will have to
be able to coin lujvo this quickly, not have to memorize place
structures systematically, and be able to learn the skill much more
easily than the thousands of words and place structures he needs to have
that 50-100K words often said characterize a typical educated speaker.

I think people are forgetting this VITAL goal when they argue for fat
lujvo, thoroughly analytical place structures, etc.

lojbab

lojbab