Return-Path: <lojbab>
Message-Id: <m0jcW7a-00024hC@snark.thyrsus.com>
Date: Mon, 13 May 91 02:11 EDT
From: lojbab (Bob LeChevalier)
To: lojban-list
Subject: Jim Carter on lujvo place structures
Status: RO
X-From-Space-Date: Mon May 13 02:11:42 1991
X-From-Space-Address: lojbab

>The ancient concept of case fits perfectly our need
>to talk about the various formal parameters (cases, places) of a
>predicate.  And it is familiar to linguists new to Lojban even if the
>more mathematical L1 terminology is not.  For these reasons I recommend
>that we say "case" rather than "place".

We do not do so for two reasons.  First is that "case" is not the word
used in Logic.  Second, "case" has some very specific meanings in
linguistics, and is a controversially defined term, with specific
denotation depending on your application.  Some theorists work with only
3 or 4 cases as applying to all of language - when we went to the
Georgetown GURP, one of the professors there chided us for talking about
"many" cases.  (Apparently Fillmore also uses a small number of cases
for analysis.)  JCB proposed 13 cases, but there was clearly some
compression.  When we started Lojban, we specifically researched case
theory, and found out that there was not only no consensus on what the
universals of case are, but we realized that the set must implicitly be
open-ended in Lojban, because in theory, it is possibly to construct a
predicate of ANY number of distinct places.  If you have N cases defined
for the language, then an N+1 place predicate will have problems.  Thus
Lojban places are not identical with "cases" in the traditional grammar
sense.

>This is sufficient justification for a campaign to interpret
>(what now are) non-predicate grammatical structures as abbreviations or
>surface structures that can be transformed to the predicate form.
>Another justification is that predicate relations are tractable
>theoretically whereas other meaning classes, such as metaphors and
>paralinguistic grunts, are so unclear as to be beyond the reach of
>logical analysis, including analysis by a machine or a beginning human
>student.

"Non-predicate constructions" are NOT (!!!!!!!!!!!) abbreviations for
predicates.  Their counterparts are not in natural language, either.
This does not mean that they may not be analyzed by COMPARISON with a
similar predication.  But they are not >abbreviations< for that
predication.  Specifically, they may either have no truth value (in the
case of attitudinals) or they may be ambiguous (in the case of tanru).
We WANT tanru to be ambiguous.  We provide plenty of methods of more
unambiguous paraphrases where a speaker wants that too. lujvo, on the
other hand, may be more susceptible to the type of analysis you want,
but I suspect that any successful analysis will occur 20 years from now,
when the language and its lujvo-making patterns are well-established,
and our equivalent of Esperanto's Kalocsay can come along and tell us
how what we have been doing intuitively can be analyzed systematically.
But such an analysis wiull be a description of actual practice, and
perhaps a useful pedagogical tool in learning already made vocabulary.
It will not be a prescription, because the prescriptive phase of Lojban
design will (hopefully) end in the very near future.

>Already it is said that all sumti tcita can be understood as fi'o <gismu>

Not by me nor anyone else I know.  This is an approximation again.  A
BAI lexeme is significantly restricted by comparison to its
corresponding "fi'o <bridi> fe'u" construct (note that this is far
broader than "fi'o <gismu>").  Indeed the latter may be so powerful as
to be useless - demanding instead a less cumbersome restrictive clause.
BAI is >approximately< equivalent to a corresponding fi'o construct WITH
ALL PLACES ELLIPSIZED.  Thus as a "predication", a BAI is SO inexact as
to defy useful logical analysis.  What we did was put in a mechanism to
eliminate the debate about what the "proper" set of lexeme BAI should
be, by allowing fi'o constructions as a "safety valve".  Thus, a safer
statement than yours is the claim that a fi'o construction can be
devised that is semantically equivalent to any sumti tcita needed which
is NOT in the lexeme BAI set.

>        la kiras. cu ja'ibi'o le falnu
>        la kiras. cu jgari binxo le falnu
>                Kira takes hold of the sail
>        la kiras. goi ko'u cu binxo le nu jgari ko'u le falnu
>                Kira changes so (he holds the sail)


Your English doesn't match your Lojban, though the Lojban is
grammatical, with the addition of a hyphen so your lujvo doesn't fall
apart into two cmavo:

        la kiras. cu ja'irbi'o le falnu
"ja'irbi'o" could be defined so as to make this mean
                Kira takes hold of the sail
though I would prefer any regularization to use the reverse order:

        la kiras. cu bi'orja'i         le falnu
           Kira      becomingly-grasps the sail.

This is justified by noting that the tanru expansion of ja'irbi'o doesn't
mean what you said:

        la kiras. cu jgari      binxo   le falnu
           Kira      graspingly-becomes the sail

since trailing places are defined by the final term of a tanru.  The
other expansion is closer, but you used an unmarked event bridi, which
cannot be VSO:

la kiras. goi    ko'u cu binxo   le nu        jgari    ko'u        le falnu
   Kira   a.k.a. x5      becomes the-event-of grasping x5(himself) with the sail.

(Note that if you intend/intended to use variations on the published
place structures, you should specifically give them when you use them
and note them as change-proposals.)

Note that we in the past HAVE done lujvo like jimc proposes here -
notably "spebi'o" for "get-married-to", so this type of guideline is not
out of the question.  But the transformation from the source metaphor
isn't as trivial as jimc would like.

Note that jimc's interpretation of "binxo" is actually closer to what I
would define "seirga'i" "self-modify".  I'll use this example to show
how >I< tend to decide lujvo place structures, and then let jimc go back
to the drawing board with his proposal, which may be quite
useful at some point if corrected (I agree with the premise of there
being fairly standard "rules" or guidelines for devising the place
structures of lujvo from the source metaphor, and indeed was influenced
strongly in this by jimc's analysis of old Loglan which is the forerunner
of this proposal.)

sevzi     x1 is the self of x2
galfi     x1 modifies x2 into x3 by doing/being x4

sevzi galfi x1 is a [self of x2s] modifier of x2g into x3g by doing/being x4g

sevzygalfi = seirgai
In determining the place structure and most logical meaning of this, we
look at the most probable interpretations of "sevzi" in the tanru, then choose
the one which is "most useful"

The "self of x2s" is a restrictive description of the modifier - a
"self" modifier, instead of a "blue" modifier or "active" modifier (such
restrictive modifications indicate no particular interaction between the
two place structures (though there may be some due to redundancy)).  The
 "obvious" place structure of such a lujvo would be:

(1)   x1 is the modifier of x2 into x3 by doing/being x4, of-type self of x5

A variation of this would be:

(2)   x1 is the modifier of x2 into x3 by doing/being x4, as a self of x5

In a sense, this may be analyzed as sticking "lo sevzi be x5" in the x1 place
of "galfi".

Next "lo sevzi" could be placed in the x2 position - the thing modified.
This gives the lujvo place structure:

(3)   x1 is the modifier of [self of x?] into x3 by doing/being x4

I've left the 2nd place of sevzi unassigned here because there are
several possibilities.  The one we pragmatically expect is that x? is
the same value as x1, and we can hence drop the place out of the lujvo
place structure to get:

(4)   x1 is the modifier of (x1-)self into x2 by doing/being x3

and we've actually lost a place.  But another, less plausible interpretation
might be some kind of mind transference operation on another, in which case
the place doesn't disappear.  I usually then move such added places to the
END of the list:

(5)   x1 is the modifier of the self of x4 into x2 by doing/being x3

because the subsidiary places of the "modifier" are usually less
important than those of the "modificand".  In this case, however, because of
the relative plausibility of (4), the following seems like the most likely
order if for some reason (4) were ruled out - by virtue of contrast:

(6)   x1 is the modifier of the self of x2 into x3 by doing/being x4

If we substitute "le sevzi" into x3 of galfi, we are talking about turning
something into a self:

(7)   x1 is the modifier of x2 into the self [of x?] by doing/being x4

Just as for the x2 substitution, there are 3 most-plausible results:

(8)   x1 is the modifier of x2 into (x1-)self by doing/being x3
(9)   x1 is the modifier of x2 into the self of x4 by doing/being x3
(10)  x1 is the modifier of x2 into the self of x3 by doing/being x4

Of course, since all of these are implausible, we can also see other
implausible possibilities spring from (7) such as:

(11)  x1 is the modifier of x2 into (x2-)self by doing/being x3

This seems implausible because of the meaning of "modify", which implies
that x2 and x3 are different from each other.  But this is an idiosyncracy
of the modifier word "sevzi". If we had been analyzing "tuple" as the
modifier, the counterpart of (11) would be:

(11a)  x1 is the modifier of x2 into (an-x2-)leg by doing/being x3

which sounds like a grisly axe-murder, but is a possible interpretation
of the tanru "tuple galfi" = "leg-modify", and shows that the form of
(11) cannot be ruled out by any rule-based lujvo-place-structure system.

3 or more new possibilities can be derived by putting "lo sevzi" in the
x4 place of "galfi" - the most plausible equating to x1 modifies x2 into
x3 just by being himself.

You might think we are done, and from the standpoint of plausibility for
this tanru we are.  But with other tanru, we might presume that the
modifier can be inserted via "lo" substitution into one of the
non-standard places of "galfi" that can be appended using sumti tcita,
such as the time, place, and comparatives and causitives, etc.  These
will USUALLY be added along with an extra term - the gismu or conversion
"corresponding to" the sumti tcita e.g.  "melbi-zmadu-galfi" rather than
merely "melbi-galfi" for "beautify".  However time/place and a few other
such modifiers may end up being added without such a "classifier", such
as turning tomorrow "bavlamdei" into "bavlamdeiga'i" =
"tomorrow-modifier".  If a Lojbanist proposed this with the meaning
"modifier on tomorrow" as opposed to "modifier of tomorrow", and
actually USED the word in this way.  I would be hard-pressed to argue
that she/had to put in a "cabna" term

bavlamdeicabga'i = "tomorrow-simultaneous-modifier"

at least partly from my intuitive limit of 4-terms for pragmatic
lujvo-making.


>In Lojban as in Old Loglan the lujvo case structures are individually
>crafted and can be substantially different from either of the component
>words.
>... this policy must
>be abandoned; you must jump over useless cases or use sumti tcita to hit
>missing ones...
>                 In practice, if the gismu cases are set up carefully
>but regularly there are few useless or missing cases, and I judge that
>the benefits far outweigh the costs of having diklujvo.

The point of this exercise has been to show just how many plausible
lujvo meanings can spring from just one tanru.  I agree with jimc that
the actual meaning (probably (4) in the above example) should be guided
by some type of predictible algorithm.  However, the extent of that
guidance should be to exclude adding in extraneous places not implied by
the tanru components, and also some kind of implication of the order for
the places that are finally chosen (it isn't fair to switch the order of
to and from places around in a lujvo of a motion word).  So I at least
half agree with jimc's above comment, as edited.

I also believe that a systematic place structure analysis of the gismu
directed at lujvo-making is beyond our capability at this time, due to
lack of person-power and the shortness of time before baselining the
language.  It is proving almost more than we have time for to do the far
simpler analysis of checking the gismu using Roget categories to be sure
that similar meaning words have matching place structures.  This has
been a low-priority task for over a year now and is less than half done.
The number of people with the time and mental stamina to stick with any
review from start to finish to ensure uniform criteria can be counted on
no hands right now.  We have pretty much accepted therefore that the
place structures cannot be baselined with the firmness of the rest of
the language prior to the 5-year usage period, which will tell us more
than any analysis will.  (Practically speaking, the momentum of 5-years
usage will make the place structures very difficult to change by dictum,
and the language may thus be less than optimal in some analytical sense.
But we will have done the best we can for our resources, and I think I
for one will be satisfied with our altogether-too-human language that
results.)

This is still only my opinion, not a policy.  We have no policy on place
structures of lujvo, nor intend to until the dictionary is started, though
I will obviously make some presumptions like the above in writing the
textbook.  Thus I welcome, and even urge debate.

-lojbab