response to jimc on diklujvo

[lojbab (Bob LeChevalier)]

jimc writes in response to Nick:
>> ...  I find gold with a metal detector, {be'i la djan.}. What did john
>> send - the gold, the detector, or me?  be'i needs a ne/pe link, and ...
>
>mi kalte lo solji sepi'o le jimkatbra  be     be'i    la djan.
>I  hunt     gold  using  metaldetector *glue* sent-by    John
>
>An aside:  See the pretty diklujvo above, jimkatbra?  Lojbab has

not pretty, ill-formed.  tb is an unvoiced/voiced cluster and needs a
'y' hyphen.  As to the aesthetics of "jimkatybra", ...

On the other hand, making this kind of mistake is an everyday occurance.
This is a correction, not a complaint.

>challenged me whether live users really interpret diklujvo right,
>without having memorized the rules. So:  I'd appreciate (I guess by
>direct reply rather than on the net) hearing whether people were able
>to interpret it.  Or, if you didn't bother looking up the rafsi, if you
>would have interpreted correctly the related tanru
>
>(jinme kalte) cabra
> metal hunt   apparatus

Your question is also ill-formed.  If you mean, does do I as a listener
understand why you used that metaphor, I certainly did.  I probably
would have used something similar myself, though more likely with sisku
instead of kalte.  But yes the metaphor is understandable.  Hmmm, though.
How about ganse, which better suits 'detector' or 'sensor'.

If you mean, would I expand it as being the equivalent of your "official
jimc interpretation".  I see several reasons why not.

The remainder of this posting is what you might call the 'long form' of
lujvo-making analysis.  You needn't do this much work for every single
lujvo you make.  Indeed the principles seem to become intuitive quite
quickly.  My claim is that jimc has oversimplified the problem in one
direction by his diklujvo strategy, while ignoring other techniques that
give better answers but are not nearly as regular.

>The official jimc interpretation is
>
>    le cabra     be lo nu        kalte fe lo jinme
>       apparatus for (events of) hunt  for metal
>
>cabra = x1 is an apparatus for function x2 controlled by x3
>kalte = x1 hunts quarry x2 for purpose x3
>jinme = x1 is metal of type x2

First of all, I will give you the benefit of the doubt and presume you
are interpreting ">le< jimkatybra", since under no circumstance would a
lujvo or a tanru be interpreted as a sumti.  I thus rephrase your
interpretation as

>       x1 is an apparatus for (events of) hunt[ing] for metal

Let me state my conclusions for people who hate to wade through my
soapish logic (lojbab, remember. %^).

Jim's interpretation can be rendered as jimnunkatybra, which has
grouping ambiguities, but makes the 'event-of' interpretation more
explicit.  The grouping can be resolved by putting in the rafsi for
'kei' between kat and bra.  But would I have chosen either longer form?
Not likely.  I just wouldn't be worried about being analytic in the
structure as jimc is. tanru are ambiguous; the lujvo built from tanru
are not as trivially obvious as jimc claims.

>Reasoning (to be carried out with a snap of the fingers): kalte is an
>actor-victim kind of word, and jinme is an object kind of word, therefore
>jinme goes into kalte x2.  cabra x2 is naturally inhabited by abstract
>sumti, so make the pre-term (jinme kalte) abstract and stick it in.

Fallacies, identified in two snaps of the fingers.

1.  Putting kalte in a tanru as opposed to se kalte, etc, or nu kalte
    is better rendered into English as 'hunter' rather than hunt.  This
    makes your source metaphor:

                 metal hunter apparatus

    This undoubtedly will immediately suggest other interpretations
    besides jimc's.

2.  To most people, jinme is a material, and hence more of an adjective
    than an 'object'  (an ambiguous term - do you mean a dacti 'thing'
    or a grammatical 'direct object')

3.  Your effort seems so analytical, what with listing out the place
    structures.  Why not spend a couple more finger snaps looking at the
    interaction of those structures and not handwave so much about 'kinds of
    words' and get a better, more thoroughly thought out answer.  That is
    in many ways the only difference between what I do and jimc does.

>If not, what interpretation did you put on it?  Or what ambiguities do
>you find lurking in the diklujvo (I can see at least two.)

Lots more than 2.  As I demonstrate:

1. Combining the two left terms:

kalte = x1 hunts quarry x2 for purpose x3
jinme = x1 is metal of type x2

gives three obvious possibilities involving the use of jinme in the
place strcuture of kalte.  Given no context to assume otherwise, we will
stick only to place structure places for this analysis, but jinme could
also be tied into a BAI-tagged non-place structure sumti place; I
consider it unfair to do so with no context, and no remnant of the BAI
tag in the lujvo, so I'll skip the possibility.

jinme in the x1 place of kalte is in effect an adjectival interpretation of
jinme - as I hinted above would seem more rational using jimc's logic above
as he worded it.  Metal hunter, in this case, is sounds like a robot dog or
the like.

jinme in the x2 place of kalte describes the quarry - this is what jimc
wants (I think).  A hunter for a metal quarry (and I don't mean a mining
pit).  This points out the next finger-snapping flaw.  Plausibility is
that a 'quarry' of hunting is living/animate, so a finger-snapping
analysis of the jimc sort would NOT likely put it here.  Now we have a
metallic quarry, and I think of the apparatus that moves a fake metal
rabbit around the dog race track.
   But of course, one then has to rule out jinme se kalte as the tanru
that you want.  If you were stopping at a two term lujvo, it would be:
jimselkalte = x1 is the metallic-quarry of hunter x2, and if you want to
turn it around to get the places in the English order, you use se again:

le  cabra     cu se jimselkalte            loi rijno
The apparatus is hunter-of-metallic-quarry Silver.

But to do this as a tanru you need "se ke":

se ke jinme se kalte

and when you realize that you are going to repeat this mess for the cabra
component, sticking it in the 2nd place of cabra, you realize that a fully-
analytical tanru/lujvo is simply not acceptible for human beings:

se <ke {nu jinme se kalte kei} se cabra [ke'e]>
selkemnunjimselkatkezyselbra (.oiro'ocai)

But this is a 'proper' diklujvo for Lojban.  If you ALWAYS made tanru this
way, they would be as unambiguous as tanru could be.  But you won't; let the
computer play such games.

Let's continue exploring place structure interactions.

jinme in the x3 place ignores the quarry (which then might as well be a
mine pit %^), and says that the purpose of whatever being hunted is
metal or metallic.  This almost makes sense.  Being rigorously
analytical would give jinme te kalte, which manipulates as the se
version to form a long-but-clear mess.

Since using either the x2 or the x3 interpretation ends up backwards
when applied to cabra, I've come up with one way to evade the problem,
in effect assuming all lujvo to be of the x1 adjectival type.

A metal-quarry-apparatus is our dog race equipment.  You can, however,
turn this around by repeating kalte, and assuming that the metal-quarry
is an adjective for the x1 place this time:

  jinme se kalte kalte cabra
  metal-quarry hunter apparatus

Of course common-sense, of the type jimc seems to want to forbid, allows
us to make this even clearer by changing the grouping to:

  jinme se kalte kalte bo cabra
  metal-quarry hunter-apparatus

Lesseenow: "jimselkatkatyborbra".  Not so pretty now, is it.  But better than
the fully analytical form, and just as clear.

The te version is roughly the same in ni melbi.

Such is the folly of being analytical.

So we fake it.  Just as jimc does, only we don't claim that we are being
analytical or 'regular'.  We assume that using a modifier in the x2 or x3
places does not require conversion, and that:

  jinme kalte

can be made into a lujvo "jimkalte" with place structure

  x1 is a hunter of metal of type x2 for purpose x3

But note that we still have 3 places.  Jimc's interpretation technique
seems to ignore the question of trailing places.  In a predicate
language, the most important semantic question is "What places are being
related?"  There is no way to say that a 3-place predicate 'emans the same'
as a 1-place predicate.  All you can do is make simplifying assumptions.

2. We now add in cabra

cabra = x1 is an apparatus for function x2 controlled by x3

Now is the metal-hunter the apparatus, the function, or the controller.
It turns out that any of the three is plausible for a metal detector.

The simplest interpretation is that it is a metal-hunter-ish type of
x1 apparatus, where we use the interpretation of metal hunter we ended up
with above.

Then we can have metal-hunter as an x2 function.  Since hunters are not
normally functions, we can presume the sumti raising that jimc does so
lightly without recognizing as such.  Nothing wrong with this - just
know what you are doing and why.

Finally, we can have an apparatus controlled by an x3 metal-hunter,
which actually sounds more plausible than the others.  But jimc has
ignored the obvious in favor of the systematic.

What place structure do I get from the sumti-raised x2 that jimc uses


                                x1 is a hunter of metal of type x2 for purpose x3
                                vv
x1 is an apparatus for function x2 controlled by x3

->

x1 is an apparatus for hunting metal x2 for purpose x3 controlled by x4

I would drop the purpose place as too ambiguous in this context.  It is
the hunting metal that has the purpose, not the apparatus.  You would
add the purpose back in using a ne/pe BAI modal attached to x2, where the
modal would most likely be tezu'e or mu'i.  Thus I end up with:

x1 is an apparatus for hunting metal x2 controlled by x3

If the device is for all metals, the x2 place can be dropped out giving

x1 is an apparatus for hunting metal, controlled by x2

3. Having done this analysis, I can now think of several lujvo that do
not have as many ambiguities as we found along the way.

How about cabra ke jinme ganse          brakemjimga'e
          apparatus-ish metal-sensor

or        cabra ke jinme sisku          brakemjimsisku
          apparatus-ish metal-seeker

The latter has a better place structure, with no purposes or conditions to
distract us:

          x1 is an apparatus-ish seeker of metal type x2 controlled by x3

We can drop the x2 out of this one as well - the result has the same
place structure but a different less glico flavor.  I'd stick with the
original order using sisku, though, to save the 'ke' rafsi, which is
vital to a listener puzzling out the meaning of that order (try
brajimsisku without knowing how we got there, and you'll have fun, I'm
sure.)

CONCLUSION:  No amount of analysis I come up with ends up eliminating the
controller place of the apparatus.  This is as it should be, because the
role of the controller is significant to it being a cabra, as opposed to
a tutci, a minji, or a zukte.  So jimc's analysis somewhere went wrong
and lost a worthwhile place.

jimc also did not recognize that he was sumti-raising from the
abstraction lenu kalte, to le kalte, again acceptible, but it should
have been noticed, especially with all the talk on the subject the last
few weeks.

The other conclusion I reach is that much of the place structure
analysis I made involved real-life experiences on the part of me the
analyst to determine what information is plausible or necessary.
Programming that knowledge into a machine is not going to be easy, and I
do not expect that we will soon have computers able to analyze the place
structures of new lujvo not in their internal dictionaries. jimc's
diklujvo might be useful as a shortcut to such analysis if it could
produce accurate results.  It did not this time.

Indeed, looking at a sample of jimc's other diklujvo gives the
impression that most are close enough to understood by someone with a
little imagination, but they are not quite optimal when looked at under
the eye of place structure analysis; and sometimes even commone sense:
"water-breathe" is a poor metaphor for "drown". jimc needs to practice
more doing it the hard way before he claims to have a universal 'easy
way'.

On the other hand, IGNORING place structures, jimc's tecnique did give a
good English-speaker's sense of what the word meant, or at least the x1
place that was used in the description of the original sentence.  I
suspect that an analysis of jimc's diklujvo will result in regularly
coming up with a useful x1, and just as regularly, garbage in the other
places.  This is suggested by his description of the diklujvo algorithm,
which uses lots of English-grammar terminology.  Only natural that his
results look like they were structured for English or other
Indo-European use; they are not suited for a predicate language like
Lojban, where ALL PLACES ARE EQUAL in determining meaning.

Now, I'll repeat jimc's question, with a different twist.  How many of
you were misled down jimc's primrose path?  Or if you didn't analyze it
yourself, how many accepted his analysis as plausible?  If a significant
number took the short cut, the danger of a process like diklujvo is
self-evident.

I reiterate the need to teach people to analyze things the 'hard way',
and let them then develop their own intuitive short cuts (I actually
knew the answer I ended up with in about 10 seconds, having looked up
the place strcutures, and I am not all that practiced in lujvo-making,
so devising such intuitive short cuts is not difficult.  Then, after say
5 years of fluent lujvo-making, we collect the wisdom of all Lojbanists,
not one malglico analyst.

Is this so unreasonable?

lojbab