Defining Lojban cmavo (and eventually gismu perhaps)

["Bob LeChevalier (lojbab)" <lojbab@lojban.org>]

At 03:52 PM 8/24/01 -0700, Nick  NICHOLAS wrote:
...
> This much is established and solid, I claim, through synergy of xorxes
> and And (and an exemplary combination of usage and rigor --- which I'd
> say combines hardlinerism and naturalism, but not everyone necessarily
> will). Being settled, it can be incorporated in a lesson.
>
> Now the outstanding stuff.
>
> Lojbab wants (and explains remarkably poorly) a protaean ka.

If I could have explained it well, we might never have had a dispute.

> And says "well, then ka is always one-or-two-ce'u, and Lojbab's protean-ka
> becomes si'o"
>
> There is much rejoicing, because Nick and xorxes (and rob, right?) now
> feel they *understand* si'o, and Free ka, and that they are the same
> thing. (i.e., the point of Free ka = si'o is that, by default, you're
> concentrating only on the essence of the *selbri*, and the actual values
> of the places are immaterial. See my parable for irreverent illustration.)
>
> PC does not understand si'o that way,

Nor does lojbab

> Nick at this point would like to invoke the {fa'a} principle. (This is one
> for the wiki, but I won't post it there right now.) For those of you that
> remember the issue, which also arose from the lessons:
>
> * Robin.TR, Nick and xorxes independently assumed fa'a (as opposed to
> mo'ifa'a) said what the directionality of the bridi event is
>
> * Lojbab (after many abortive attempts at communication) said no, it's
> position, like all the other FAhA
>
> * At this point, Nick yields, compelled by the analogy with the other FAhA
>
> * Someone (Jorge? Nick?) says "but how are we supposed to know what it
> means?"
>
> * Lojbab (and I'm sure he's regretting this now) says "Well, John and I
> know what it means."
>
> * xorxes calls him on this. "No. If you're the only people who know what
> it means, and don't tell anyone else, then everyone else will work out on
> their own what it means, and their usage is what will prevail."

At this point, my kids being back in school, and Nora off for a week of 
vacation (during which she plans to finally make serious headway on 
reviewing stuff of Nick's that needs her review), I will introduce the plan 
we started to talk about around Logfest to answer this.  I hope Jay is 
paying attention, since it seems he is likely to be most able to provide 
the net support for this.

What is now the refgrammar originally started as a set of example sentences 
for every cmavo.  Cowan found the job too difficult because we weren't even 
all that solid about how the grammar at the selma'o level would work, and 
instead started to write a selma'o catalog.  That catalog is now but one 
chapter of the refgrammar, and the real meat was integrated topically into 
the other chapters of the book.

Now it is time to return to the original effort, which it now seems is 
necessary (per the above discussion) in order to have a minimally useful 
cmavo dictionary as part of the official baseline dictionary.

The goal is to have example sentences for each cmavo IN EACH GRAMMATICAL 
ROLE IN WHICH IT  CAN HAVE DISTINCT SEMANTICS.  Thus, if we were to agree 
on conventions for ka with and without various ce'u, zo'e or whatever, each 
conventional usage would be recorded in example sentences of each of the 
different types.  These example sentences might need an English translation 
and also might need a (hopefully brief) English language explication 
especially if the English translation is semantically ambiguous.
In addition to conventional usages, we also need usages that are 
grammatically distinct.

Thus we need an example of fa'a per Nick's discussion above, and we also 
need an example of mo'ifa'a.

Per the issue raised by Pierre, we would need an example something like
> ko'a klama .isecaubo mi djuno

as well as the more likely
ko'a klama secau lenu mi djuno [tu'ala'edei]

as well as
ko'a ne secau mi [cu] klama le zarci

Contemplating this effort, while I'd love to be able to do it all myself, 
just as Cowan largely did the Book all by himself, and thereby assure a 
single voice nominally producing a coherent set of definitions and 
examples, it is clear that I don't have the time, and perhaps not the 
Lojban skill and creativity to produce such a catalog.  Cowan seems 
unlikely to have the time either.  It really isn't a one-man job.

Luckily we are not a one-man community, nor do I feel any need to reserve 
to myself the sole power to decide everything about the language (per the 
"lojbab lesson" that I think Cowan has referred to a few times in recent 
weeks).  Maybe we will still need a Lojban Central to resolve any hard 
cases where multiple conflicting examples are created, but maybe even those 
issues can be resolved sekai leka jbocecmu (perhaps an example of a 
ce'uless ka.  No doubt ce'u belongs somewhere within the ka bridi, and 
perhaps in multiple places, and in all likelihood at least some of the 
places a ce'u might be considered are BAI or FIhO places so that there is a 
potentially infinite number of ce'u to consider, But I am not considering 
them individually, so I want not to have to specify  ce'u that would 
mislead - it is the community-ness as a whole that someway relates to how I 
would like to see decisions made about the language. [If this is worth a 
side discussion, let's move it out of any thread discussing cmavo 
definition writing.])

I would not be satisfied with a Wiki-style approach to this.  I want to 
constrain what is submitted so that discussion and argument in the 
definitions is minimized - we want Lojban examples.  I want the 
traceability to individuals and something like the version control that CVS 
provides, but I'd like it to be more spontaneous -such that someone who 
thinks of an example for a cmavo can go to a web page and enter it then and 
there without checking it out of a library.  Perhaps it is sufficient that 
examples and minimal explanation can be added by anyone but deleted only by 
a select group that could edit multiple submissions into a cohesive unit (I 
don't propose that this editing be done on any substantive issues until we 
have a substantially complete set of usages, but we need some way to delete 
or correct erroneous Lojban, typos, etc. in short term, and people will 
thus want to enter correction notes as commentary, and the editors will 
then go back later and make the corrections.)

The exact format of data entry is up for discussion and of course to 
considerable extent the discretion of the person doing the software 
support.  It is possible that Jay's ongoing dictionary effort already could 
support this, or would be able to do so with minimal changes.

The bottom line is that we would end up with examples that would supplant 
the need for relying on Lojbab's brain as to what the meanings of the words 
are.  I would be as obliged as anyone to submit examples of usage if I want 
meanings to be established.

Since this is intended to become part of the baseline dictionary, we 
probably need to start by capturing what can be captured from the sentence 
examples in the refgrammar.  But if the refgrammar is ambiguous about the 
meaning of ni, then I believe that the dictionary has the power to 
prescriptively resolve that ambiguity, and indeed with respect to semantics 
can supplant the Book to a limited extent because we recognize that 
semantics is the province  of a dictionary more than of a refgrammar.  (I 
believe we actually have a resolution in writing somewhere as to the 
relationship between the various baseline documents in the event of 
contradiction).

I also would like to capture examples from actual Lojban text or 
translations, and we would value those more than ad hoc examples (because 
"usage decides").  The dictionary data base needs to allow for entry of 
reference citations when an example text is the source - URL citations may 
be sufficient, but the references won't all be on lojban.org, so we need to 
then ensure that copies of all cited documents are collected in a single 
archive site so that the citations have meaning in the indefinite future as 
Web sites appear, disappear and move.  Since I would prefer that debate on 
semantic issues be NOT included, just decisions, there should also be a 
means of citing to the pu'o Elephant, the wiki, or the List in the 
explanations, and this should be distinct from a usage citation.

(Not knowing exactly what is being done for Cowan's Elephant, a second part 
of the LogFest proposal was that semantic issues and resulting decisions 
should be recorded formally when they are actually as decided, not quite as 
pc has done so in the past (though his effort was part of my inspiration), 
but rather in the form of Cowan's Change Reports for the baselined 
grammar.  These Semantic Reports would include a statement of the issue 
including examples showing how confusion exists, rationale for the multiple 
approaches, and then a conclusion if one can be reached.  I had intended to 
develop this idea further by now, and also to have examples that I asked pc 
to generate, but neither has happened, and depending on how the Elephant is 
designed, it may incorporate what I am seeking.)

I do not feel the compulsion that all decisions of usage have to be 
back-compatible with all past usage, something that I fear seems to be an 
extreme referred to in the limited discussions I've had time to read.  Thus 
Nick does NOT have to revise all of his writings in order to ensure that we 
don't insist on weighing his copious examples as constraining later usage 
made with more understanding of features that weren't understood, and 
perhaps not even in existence, back in 1991.  If there is no better 
example, then a 1991 usage may make it into the dictionary, but people who 
think that such a usage is wrong are free to create new usage and add it as 
an alternate example.  Perhaps they will need to write a longer text in 
Lojban just to establish context allowing the specific example they wish to 
make.  But then they have written a Lojban text (which can be added to the 
citation data), which has greater weight than an ad hoc example that has no 
context.

So the steps I see are:

1. Set up the software support to include the appropriate data base and web 
based forms and editing support (and perhaps some utility software to allow 
bulk submission of off-line work in some prescribed format).  Useful would 
be a way that people can display what has been done to see what still is 
needed, and also to have the draft effort such as it is serve as a better 
cmavo dictionary than the official cmavo list.

2. Transfer into the data base the information in the baseline cmavo 
list.  We need to use my off-line master, which I believe incorporates all 
the cmavo that R. Curnow asked about, even if they aren't on the web site 
for some reason.  (Maybe I'll get to fixing that as part of turning over 
the lojban.org site management to Robin.)

2a. Still to be determined is how to deal with cmavo compounds, some of 
which exist in the cmavo list.  cmavo compounds are in the final analysis 
merely one more kind of contextual example of usage, but we need to decide 
which cmavo compounds rate cross-reference entries.  My default assumption 
at this point would have an example for mo'ifa'a be entered both under fa'a 
and under mo'i, with an added cross-reference for the compound that would 
point to the component words.  This seems straightforward for a web 
dictionary, but figuring out how to deal with this in the paper dictionary 
is still up in the air, and may be until after we have many example entries 
of these compounds.  Note that the number of compounds is open-ended, 
unlike the number of cmavo (after all, there are an infinite set of numbers 
that are all cmavo compounds). Thus the dictionary data base has to be able 
to support adding entries that are compounds, but not adding cmavo (see 
below re experimental cmavo).

3a. Add example entries with citations for all sentence examples in the 
refgrammar.

3b  Independently, people can be adding entries based on other text 
citations, and ad hoc entries.

4.  Once 3a is done, there may be enough data and examples that we might 
see some changes needed in the data base and processing.  So temporarily 
freezing the data base, make any necessary changes and go back on line.

5.  When we have examples for most or every cmavo in every grammatical 
context that seems necessary, an editorial group can start coalescing the 
submissions into a cohesive unit.  Who is on the editorial group will not 
be decided until we need to decide; this is not the current Lojban Central 
taking over everyone else's good work.  At that point, we will also 
determine how to resolve contradictions among the submitted examples.  That 
will be the point where someone (Lojban Central or whoever) may finally 
answer whatever questions about semantics will be answered prior to baselining.

6.  When the bulk of the cmavo work is done, it would be useful to create a 
similar effort for gismu and perhaps even some lujvo, so that questions 
about the meanings of gismu can be resolved by resort to actual usage 
examples.  I don't think that we will understand what we are capable of in 
this kind of effort until we've done the bulk of the cmavo entries.

Experimental cmavo are a problem.  If they are official recognized in a 
baseline document, they aren't entirely experimental, and they to some 
extent acquire the stature of the baseline.  Some of the xV'V cmavo have 
been used multiple times in proposals to ad this or that feature to the 
language, and because the usages are experimental, there is no requirement 
that anyone respect any one else's experimental usage (they may want to and 
they may want not to do so).

As such, while I see no problem in building a data base of experimental 
cmavo usages, citations of usage, as well as the interface to add to and 
eventually to edit those usages will likely be identical to the one used 
for the official work, I want the on-line data base presentation and entry 
for experimental words to be distinctively separate from the official words 
so there is no confusion (if people can add to the experimental cmavo and 
add to the compound cmavo, but cannot add to the cmavo set itself, this may 
be sufficient separation on the data entry side, but any displays and 
lookup function should not mix the two).

> Nick ardently wants {fa'a} to be about location, to maintain the cohesion
> of FAhA. Likewise (being a hardliner) he'd rather a Dictionary
> fascistically settle these things, whether by example phrase or logical
> formula --- or preferably both.

I want to ensure that there be example phrases.  Whether someone wants to 
go through the trouble of adding logical formula where applicable is 
probably up to the potential adder.  If you have in mind a particular 
format, Nick, I'd like to see you pose an example of a cmavo usage, and 
example phrase and a logical formula of the sort you'd like so that Jay or 
whoever designs the data base in a way that supports such, with the formula 
distinct from commentary if that is appropriate.

> But on {si'o}, he calls out pc: if pc is the only one who knows
> what {si'o} is, then other Lojbanists are going to go ahead and construe
> {si'o} how they want, and their usage will prevail. You now have two or
> three Lojbanists who say "Aha! So *that*'s what {si'o} is about!" This
> genie is unlikely to go back into the bottle.

This effort is a way to at least put multiple usages in the record so that 
it is possible to intelligently resolve a dispute among multiple usages if 
such multiple usages are presented.  I think I agree with what I understand 
as pc's position, but the proof will be in the examples.  If over a long 
period of time (as opposed to ad hoc in a list discussion when traffic is 
running at hundreds of messages a week), no usage examples surface in 
support of pc and my outlook, then usage will indeed have decided.  If 
contradictory usages are entered into the data base, we'll decide later how 
to deal with it.

> Nick contends (and I think I'm echoing Adam here, but I'm not sure) that
> free {ka} is bogus, and that Lojbab is confusing "property" (of specific
> places) with "quality" (which concentrates only on the selbri, I claim).

Perhaps.  I haven't thought much about the meanings of the English words 
"property" and "quality".  I associate ka equally with those words and the 
English suffix "-ness".  Several place structures in the gismu list as well 
as the definition in the cmavo list of ka associate it with both "property" 
and "quality".  Some of the gismu places that refer to "quality" are such 
that it is not clear whether some particular value or perhaps multiple 
values are to be referenced via ce'u within the ka abstraction that fills 
the place.  Since there never was any effort NOR ANY INTENT to define most 
words of the language rigorously prior to usage (because I was afraid that 
a) we couldn't do so coherently and b) to do so would be to impose 
metaphysical constraints on a language that is trying to remove such 
constraints wherever possible - and this is especially dangerous while we 
are predominantly English speaking lest malglico leak in), it is difficult 
for me to argue with you on this, and I would rather not do so.

As I have noted in another post, attempts to equate ka and du'u and si'o 
etc, with each other seem inherently to be imposing metaphysical concepts 
on the language.  I'm not prepared to concede particular relationships 
between properties, qualities, ideas, experiences, expressions and 
propositions that are not mandated by baselined place structures because 
they so impose those concepts.  We are better off with vagueness, and the 
risk that 3 people will *unofficially* railroad a usage that I don't think 
I agree with into such common usage that the "genie cannot be put back in 
the bottle", than we would be with an official ruling that we might later 
regret because it constrains later usage as only official baseline 
statements do.

> Once {ce'u} was introduced into the picture, he
> contends, {ka} is about properties, not qualities.

That is not what the cmavo list says, nor what the gismu list says when it 
refers to qualities and ka in the same place, and it is not a requirement 
of the *grammar* that ka be solely about ce'u.  That is a usage issue; it 
can be left to usage.  Just as the lujvo-making place structures are 
strictly a convention (in this case an officially recognized one) there is 
nothing prohibiting a violation of the conventions proposed regarding ce'u 
so log as they are kept unofficial, and given the existence of the gismu 
list referring to ka as property or quality it would be wrong (i.e. 
baseline violation) to try to so prohibit.

More importantly, I am somewhat resentful that, having decided that ka 
cannot be about qualities, people are trying to say that si'o or du'u or 
whatever IS about qualities, and not about what they were described as 
being about in the cmavo list.  I will admit that things have been unclear 
in documentation about exactly what these words are supposed to mean.  Mea 
culpa, I've only been promising a dictionary "within a few months" since 
1988. %^)  So I concede that it is about time that we finally get things 
down in text as to what words mean, and try to reach resolution where there 
is debate and where such resolution can be made.  (Unlike you, I am willing 
to accept the possibility that some words will remain under dispute as to 
their semantics even after the baseline dictionary is written, but I am 
willing to accept that if there is such dispute, that we should document 
the nature of the dispute through examples and if necessary explanation as 
part of the baseline documentation.  It is fair to ask that the baseline 
document what is NOT decided as well as what IS decided.

>  Qualities are {du'u},
> or {si'o}. Free {ka}, like bound {ka}, should by default be assumed to be
> a property of just one thing.
>
> If, as someone here constructively
> suggested (maybe Lojbab, maybe Rob), you want to have a special
> "metalinguistic mode of discourse", in which {ka} *become* by default
> qualities (all-ce'u) rather than properties (one-ce'u) --- then go ahead.
> That's the thing about pragmatic conventions: they're defeasible. But
> since many other Lojbanists will see {ka} and think property rather than
> quality, it's your responsibility to make the reader realise that, where
> you're leaving {ka} places blank, assume they're *all* {ce'u} in this
> text. (Or, put all the ce'u in, as others have suggested.)

If they are official, it is harder to ignore a convention.  I am resistant 
to official pronouncements on semantics that are non-trivial.  I think I 
understand why you feel otherwise, and I'm not sure how to resolve things, 
other than to document multiple points of view to the extent possible and 
then let the users, if not the usage, decide.

> Nick also sticks by his simplistic (but used in the Lessons) dichotomy:
> {nu} is something that happens (or may happen) in the world; {du'u} is
> something you hold in your head.

That presumes that all things that happen in the word are expressed in nu 
and nu along, and all things that happen in your head are expressed with 
du'u alone.  I don't agree with this dichotomy, though I am probably closer 
on the nu side than I am on the du'u side.  I would be more inclined to 
accept a dichotomy between su'u as a mental thing and nu as a happening., 
with du'u being one type of su'u and possibly therefore excluding nu from 
su'u.  But the problem I see is that this insists on a metaphysical 
dichotomy between what happens in our heads and the real world, and that is 
not something I want to impose on the language.  In particular, ideas and 
experiences straddle the boundary - they tend to start in the heads of 
individuals, but they take on a life of their own.  Given that we observe 
the real world only through mental processes,  I can easily see a 
justification for refusing to separate the real world from those 
processes.  BUT THE POINT IS THAT WE'VE DRIFTED INTO METAPHYSICS - this is 
not language any more but philosophy.  Let us try to allow in the language 
for metaphysical ideas that are not in accord with early 21st century 
thinking even if we find it hard to come up with ways of seeing those 
alternatives that we don't necessarily subscribe to ourselves.

I thus have no problem with your teaching such a dichotomy in your lessons, 
so long as (and this is just a suggestion, since you have editorial 
control) you include somewhere the disclaimer that this way of looking at 
things reflects a particular way of looking at the ideas represented by the 
words, and need not be the only way (or some othet metaphysical bias 
disclaimer).

> (I don't give a damn whether the {du'u}
> is truth-conditionally propositional or not. This may be an error of
> mine.) To his simplistic way of thinking, if a {si'o} isn't something you
> hold in your head, he doesn't know what is.

A si'o is something that probably starts off held in the head of one or 
more tersi'o, though different tersi'o may come up with the idea or concept 
based on different events or properties or qualities (selsi'o).  But ideas 
can be transmitted, shared and adopted, while experiences can merely be 
described and empathized with, and qualities have a theoretical existence 
whether or not any mind bothers to think about them - they never belong to 
a thinker at all.  I'm starting to see a place structure distinction here 
that you may have missed or rejected %^).

In any event, were one to explore the metaphysics that the "real world" is 
just the dream of the butterfly Lao Tse, or perhaps that some or all real 
world events are manifestations of the mind of God, your dichotomy clearly 
breaks down, and indeed could interfere with communication.

> Nick also now thinks the battle over Free {ka} vs. {si'o} is not as
> important as the battle he lost to Adam, when he (I) was trying to
> conflate Free {ka} and Bound {ka} (by filling places.) I claim that we have
> enough consensus about Bound {ka} --- the really important {ka}, the one the
> gismu list *forces* you to use --- that he can write something
> constructive in the lessons.

Only in reference to the place structure of ckaji, which I admit is the 
source metaphor for ka, and for some gismu based on similar concepts.  It 
is possible that in most place structure usages that call for a ka 
abstraction, there is implicit ce'u, though at a glance, terzu'i and 
terdicra might be ce'u-less in the same way that Lojban distinguishes 
between agents/agentive causes and physical causes (i.e. there may be no 
clear way to fit le dicra into a particular ce'u place of le terdicra).  It 
is also possible that some metaphysical problems may exist that should 
negate even this concession on my part.  But without having an example to 
argue with you from, I'm not going to pretend like I have one.  Just be 
prepared to have one jump out and bite us when it is inconvenient.

Meanwhile [ka bridi] is grammatically a selbri or even a tanru-unit, and 
not a sumti of one or more gismu.  As a selbri on its own, as opposed to 
its usage in "leka bridi", it is harder to classify ka as solely a creature 
of ce'u.

lojbab
--
lojbab                                             lojbab@lojban.org
Bob LeChevalier, President, The Logical Language Group, Inc.
2904 Beau Lane, Fairfax VA 22031-1303 USA                    703-385-0273
Artificial language Loglan/Lojban:                 http://www.lojban.org