Message-Id: <199311182236.AA00336@eli.CS.YALE.EDU>
Date:         Thu, 18 Nov 1993 17:31:31 EST
Reply-To: jorge@PHYAST.PITT.EDU
Sender: Lojban list <LOJBAN%CUVMB.bitnet@YaleVM.YCC.YALE.EDU>
From: Jorge Llambias <jorge@PHYAST.PITT.EDU>
Subject:      Re: response on old posting - lujvo-making TECH/PHILOSOPHY (long)
To: Erik Rauch <rauch>
Status: RO

ranji fa le nu casnu le terjvo

> Responding to Jorge's comments on my long posting:
>
>  The concepts exist in natlangs, but when
> you are thinking IN LOJBAN, you will hopefully not be phrasing the
> concepts in other-lang terms before phrasing them in Lojban.

I agree. At that point one can be said to have learnt the language.
I am able to think in English in an English context, and in Esperanto
in an Esperanto context. Although I can more or less read French, I can't
think in French, and I definitely don't know enough Lojban yet to think in
Lojban.

> As a
> result, you may not grab hold of the same component gismu that you would
> when being fullt analytical based on transaltion.  You will probably, if
> you think of a goer, a route, and not a destination, grab hold of litru
> rather than klama, even if in English or Spanish, you might express it
> in words that strongly connote origin, destination, or means.

I don't understand why you would pick klama when being fully analytical,
but I agree that litru should be the right choice, so that you don't have
to drop the origin and destination places arbitrarily. However, once you've
picked litru, you still have a means place, want it or not. What I say is
that if your concept doesn't have a means place, litru would not be such
a good choice for that lujvo, or you'll have to modify your concept
slightly.

> But that is the point.  There is NO single Lojban expression that is THE
> translation for "dog", and you need a context (under conditions) and
> standard place to talk about "better" translations.

Of course, this is true of any two languages. I was talking of a particular
instance of the use of "dog". I agree that we shouldn't be looking too
much into the meanings of words in natlangs, also.

> In any case, whether the English normally does or does not include the
> se gerku is irrelevant to the Lojban lujvo - and indeed I contend that
> what matters when you invent a lujvo using gerku, is whether your
> concept DOES have a significant use for the se gerku.

Since I think that the x2 of gerku is already superfluous, I will agree
that it does not belong in any lujvo involving gerku. In these cases,
the problem I see is in the source gismu being fat, not the lujvo being
fat as a result of not allowing deletions. I think that this is not the
case with most gismu. I think most of them are not fat, even though I've
been complaining about a few.

Let me give another example from the jvoste, also mentioned in Nick's paper:

jditadji        policy: t1 j1 j3 t3
x1 [process] is a policy (method of deciding) for x2 (person) about matter x3
under conditions x4

jdice     jdi      decide
x1 (person) decides/makes decision x2 (du'u) about matter x3 (event/state)

tadji              method
x1 [process] is a method/technique/approach/means for doing x2 (event)
under conditions x3

We don't want the se jdice as a place of "policy", (although I can think
of meaningful uses with kau) so we throw it away. I would prefer to leave
the place, and use for example

pairtadji       t1 (t2=nu pajni) p1 p2 t3
x1 is a policy (method for deciding) for x2 about matter x3
under conditions x4

pajni         pai  judge
x1 judges/referees/arbitrates/is a judge determining/deciding matter x2


Of course, we could have the two as near synonyms, but I find the
disposing of a place in jditadji arbitrary. I agree that a concept
exists without that place, but I don't think it should be represented
by that lujvo.


> If it does not,
> in the concept you are considering, I do not agree that you need to
> explicitly mark the deletion with zi'o into the lujvo.

I too would prefer zi'o not be used in lujvo.

> Now someone else
> making a similar concept at a different time which DOES rely on the se
> gerku place, might get a different place structure.  This will happen,
> and eventually consensus will decide,

Agreed. I'm not opposing the place structure that develops after extensive
use. What I don't like is that a word appear in a dictionary after
coinage with an arbitrary deletion of places based on a single instance
of use (or a few instances by the same speaker) to fit a "concept" that
is not what the components would suggest.

> Furthermore, to minimize our gismu and maximize their lujvo-making
> power, we have tried to abstract some of them a little beyond English
> norms (not applicable to gerku, but true for others, as Colin often
> points out).

I'm not sure what this means. I would think that the more places a gismu
has, the less abstract it becomes.

> Presumably, on the other hand, most lujvo, especially lujvo of 3 or more
> terms, will rarely be used as building blocks, and will represent less
> common concepts.  Less common concepts tend to be more specific and
> concrete.  When you have a concrete concept in mind, then various
> abstractions present in the building blocks used to express that concept
> may be irrelevant to the specific concept you have in mind.  Now if you
> are taking time, and wish to analyze things, out, you will look at all
> such places and try to envision whether you can imagine they would EVER
> be relevant or metaphysically important.  In scholarly written Lojban
> this might be more common in lujvo-making.  Hopefully also in poetry.
> But not in informal or spoken Lojban.  And since we hope that the latter
> will be the norm, we have to expect that most words will be added in that
> mode, and recognize what effect this will have on lujvo-making.

Given that when using a lujvo many places will be left empty anyway, I don't
see how we can tell whether the place actually is a part of the lujvo
or not until it has been used a great many times. Whatever the first
use of a lujvo is, it's not really a part of the language (as a word with
its own meaning) until it has been used a lot. I don't suppose lujvo-making
will be a significant part of Lojban speaking once commonly used lujvo
exist for most commonly used concepts.


> >Ok, so you say that it's legitimate to use a rafsi to represent either a
> >gismu, or a zi'oed gismu.  I don't like this.  I'd say that the
> >likelihood of that some other gismu to exist should be proportional to
> >the likelihood that it will be needed.  The better the gismu cover
> >"concept space", the more likely the gismu you need will exist.
>
> But that was not the basis used for deciding what gismu there would be
> in the language.

I thought that the two thousand or so gismu were selected on frequency
of use in natlangs. Since most concepts can be expressed in simple
(ie often used) words, most concepts should be expressible using gismu.

> >The point is that such a lujvo won't become a dictionary word unless it
> >enters the language by being used (more than once and by more than one
> >speaker, of course) or it is written in the dictionary as a definition
> >(like all gismu).  In this last case, I don't think we can simply remove
> >places because they don't fit the concept.
>
> I'm hoping that this latter case will be the exception rather than the
> rule in adding words to the dictionary.  Thus far it is. the 3000-odd
> lujvo that are being added so far include mostly lujvo created for a
> specific usage and context, as opposed to the Eaton effort.

I would include lujvo created for a specific usage and context in the same
group as those for the Eaton list. The other group would include those
lujvo that have been used, say more than one hundred (or whatever) times.
How many of those 3000 lujvo have appeared in more than one text?
And how many have appeared with all their places filled? Are the places
that appear in the dictionary only the ones that were filled in the texts?

> The habit I want most is to THINK ABOUT the component gismu, and where
> possible, their places considering each for relevance, and not
> willy-nilly including gismu based on their keywords (which is the
> typical alternative).

Of course I agree with this. The point at which we depart is what to
do when we find that a place doesn't fit the concept we had in mind.
You say remove the place, I say look for a different lujvo (or accept
a modified concept).

> If we do this well - just this one rule - we will end up with a far
> better language for neutrality purposes than any other single rule, and
> I suspect we might end up with a better overall language than adding in
> a detailed set of rules that might trade spontaneity for
> accuracy/predicatability.

I suppose it depends on what you mean by doing this well. As long as the
language becomes a living language, it will serve as a language. If the
lujvo are more or less regular, that will affect how easy or hard it is
to learn them, but I don't think will change much the language's
"neutrality".

> I also want to minimize the necessity for people to memorize place
> structures thoroughly or accurately, as part of learning the language.

The danger of making lujvo without knowing what are the place structures
is that one is not using the real meaning of the gismu, but some meaning
related to the keyword. Of course, you don't have to remember the order
of the places, but to know what a gismu means you can't avoid knowing
what its places are, in theory.

> Skilled Lojbanists will probably know most of the place structures, but
> the typical learner won't, and likely won't even know every single
> gismu.

So gismu created by the typical learner as they learn are to have the
meaning as they are first used, until someone decides to use it with a
different meaning? I'm not saying learners shouldn't create lujvo
(it would be a bit hypocritical to say that), but to what point should
those lujvo enter a dictionary, and with what place structures?
(I don't care too much about the order of the places, but yes about which
places are present, which determines the meaning of the lujvo.)

>  Once we have a large body of
> fluent speakers, they may look at these initial lujvo and indeed think
> them atrociously irregular and set in some more systematic word-making
> rules,

Once there is a large body of fluent speakers, all the irregular lujvo
that have entered the language will remain a part of it. It is very
difficult to change the meaning of a word once many people use it with
that meaning. I'm not saying this is bad. I think it is inevitable,
no matter how much we try to make things regular, irregularities will
always appear in a human language.

> and Nick's effort seems like a more than reasonable first cut for
> such systematization, and an effective way to set initial place
> structures for a lot of words where we simply don't have time to work
> out place structures more 'naturally'.

The only way I can think that would be more 'natural' is not to record
any place structures until they've been used a great deal of times.
I don't see why selecting some places to drop is more natural than letting
them stay.

> Of course, I worry that large numbers of Nick's lujvo, especially after
> the weeding he has done on them, and after applying the place structures
> he has algorithmically defined for them, will not work properly in the
> texts from which they were originally culled.

Maybe we could look at some examples? You mean that they wouldn't work
because of reordering of places (surely this is unimportant) or because
of places being added and/or deleted?

> And I have agreed that Nick's approach is better than any other idea we
> have come up with.  I just want it considered that these rules might
> ONLY be appropriate for adding words in dictionary-making mode, and are
> not the best mode for analyzing real texts

Yes, I guess we're talking about how to make lujvo, rather than how to
understand lujvo made by others. I don't think making a lujvo for a
text should be so different from making it for a dictionary. As for
informal lujvo-making, that's perfectly fine, but I suppose those
won't make it to the dictionary unless they've been used very often.

> (I would hate to see a new
> Lojbanist composing their first Lojban text, including several lujvo
> coinages, and getting comments back based heavily on Nick's scheme.  It
> would cause me, for example, to give up trying to write in the language.

The tact with which beginners' texts are treated has little to do with
how to determine place structures, I think. Personally, I prefer that
people react to my texts and show me where I made mistakes, or could have
used a better expression. This way I know that someone is reading, and I
can always defend what I think is still right, or make the corrections and
learn when I made mistakes, or accept suggestions for improvements.

> On the other hand, for trying to get a good Lojbanist like you, Nick, or
> me up to a more skilled level, such analysis is worthwhile for critique,
> as long as the author retains the right to let their instincts override
> the rules on any particular lujvo.

The author should retain the right to use any lujvo with the place
structures they please. The question is whether the community will pick
up such lujvo and use them as the author did initially. That will of
course depend a lot on who is the author, and that's why I think it
is worth trying to convince you not to drop places just because they
don't fit a preconceived concept, since people will tend to accept
the lujvo you and the more experienced Lojbanists create. Since it is
almost impossible to determine just from usage the structures of most
lujvo created up to now (most have been used only once, or in only one
text), there has to be some way that the dictionary maker decides what
places are part of a lujvo and which are not. Also, any lujvo that appears
in the dictionary will tend to be fixed with that place structure, since
people will tend to use it that way.

> >> I think people are forgetting this VITAL goal when they argue for fat
> >> lujvo, thoroughly analytical place structures, etc.
> >
> >I haven't seen those arguments, so I can't comment.
>
> I apologize if I have falsely characterized your views.

Well, not falsely, just no'e satci :)

> I contend that
> a policy that lujvo contain all places implied by component gismu UNLESS
> there is a specific rule-based deletion transform is a policy for fat
> lujvo and analytical place structures.

It is not thoroughly analytical, because it still leaves lots of room
for place ordering, which places are doubled up with which, etc.
You could also say that a policy to allow places that are absent in the
component gismu as places of the lujvo is a policy for fat lujvo...

> ... on the other hand, I can imagine that as we get more skilled at the
> language and lujvo-making, we will come to recognize new patterns in the
> types of deletions that we want to make in place structures that will
> make the deletions seem less arbitrary, as in this case.

Great! Let's get more skilled quick then!

> We can hope so.
>
> lojbab

Jorge