Re: response 1/14/92

[cbmvax!uunet!math.ucla.edu!jimc]

Nick responds (1/25) to Lojbab's message of 1/14.
> >John continued to defend your position on dikyjvo as he interprets it, which
> >not quite what I think you say, and certainly not what Carter wants.  He see
> >to water down the position to near-meaninglessness.  His interpretation, as 
> >understood it, is that you believe that dikyjvo should be guidelines not rul
> >and that they should apply to people trying to decide the place structure on
> >they know what the word means, if they can fit it to one of the patterns.  
> 
> While regretting that Carter's quixotic dream of an unambiguous lujvo making
> algorithm is unrealisable in Lojban, yes, this is, I believe, the main benefi
> of dikyjvo, and how I end up using them. I am aware of them, but I don't let
> them dominate; if a rearrangement of sumti has to be made, I judge it on the
> spot. The point is that a lot of judging of place structure (as opposed to
> meaning, or dikyjvo flavour - as I pointed out with zasyspo, the two are NOT
> equivalent - something Jimc might well pay attention to) can be automated
> and regularised and rendered predictable - which I believe is great. I know
> that having to judge every place struct. on the spot instills one with a grea
> *horror vacui* - you don't know where to begin.

Quixotic, eh?  Well, it's true that dikyjvo are grafted onto Loglan-Lojban
after the fact, and hence there is a big problem determining the flavor
of the dikyjvo.  I prefer to fiddle with places and even with syntax
to make rules possible.  But to my mind, if rules are impossible
then guidelines are a *whole* lot
better than nothing.  And as Nick points out later, some categories of
dikyjvo are so obvious and so regular that even if defined per 
guidelines, people will treat them as rules.

> It's better than judging them each individually ground-up, which seems to be
> current LLG policy. There are some lujvo whose structs. simply need not be
> controversial (though with my {fai} prtoposal above, this remains to be seen)
 > .
> I don't find it weak - when I come up with forms like {fuzysku} in Aesop,
> I feel empowerment - of a substantial form, not the "potential" which you've
> described yourself as emoting (pu'i, not ka'e).

Yes!  It's not reasonable to ask the users either to hand-craft their
own lujvo on the fly, or for LLG to hand-craft zillions of lujvo and to
pass them out in a dictionary.  When you make a rule/guideline you
automatically make a very large number of lujvo, and these will actually
be useful for the speakers.  

"Empowerment" excellently expresses the feeling I get when I use dikyjvo.
I can express the meaning I *want* and I can expect the listener to
understand it, without having to go back to Lojban Central to get a
word built.

> >The contention was made (by John) that you and Carter are merely extrapolati
> >from actual patterns used by JCB, and by me and you etc. - i.e. it is in its
> >mostly descriptive.  I reject this claim because I dont; think anyone beside
> >you has used lujvo extensively enough to even have a statistical pattern d
> >develop

Sort of, sort of.  The first step was to analyse the lujvo in the Old 
Loglan dictionary -- not statistics weighted by usage, but by number of
separate words.  There wasn't any usage at that time.  
The second step was to refine the categories and
"flavors" into something comprehensible, and to propose a moderate
number of place structure changes which would minimize the number of
flavors and would make the remaining flavors more general and useful.
Finally -- and this was in -gua!spi -- I ensured that flavors could
be resolved by syntactic means (not available in Lojban), and I revised
place structures wholesale, to further simplify the rules, plus I 
inserted replication rules into the definitions.  

Thus the Carter dikyjvo rules began on a foundation of JCB lujvo, but
were considerably modified (simplified) through feedback from Carter
usage (whose else could I analyse??).  A key process in the development
of the rules was interplay between the combining rules and the place
structures of the words that "wanted" to combine, so that the rules
generated the lujvo desired by the user (note singular number :-).

So someone being churlish could say that I made the whole thing up
from thin air and a little JCB trash.  But the patterns revealed are 
so obviously useful...

>  What Jimc did which
> is *substantially* important is the {belenu} analysis, of {ri'a} and so forth
> which, at least for {ri'a}, has a very well def'd place struct. When applied
> to other gismu, as I pointed out, you get replication and other such compli-
> cations. And because the transformational mechanism of {belenu} dikyjvo is
> unique to Lojban, this flavour can be said to be non-descriptive of NLs, but
> rooted probably in JCB's usage. It had to be pointed out to me before I could
> run with it. And the prescription doesn't fit everything I want to say. But
> precisely because its place structs are well def'd, this flavour of dikyjvo
> is the most important and productive, as well as the most alien.

That's a real interesting comment.  It is certainly true that the 
abstract dikyjvo make the most words, the most useful words, and the
ones which are most prolix when written out in full.  Having been close
to the topic for a long time, I hadn't realized, but Nick is quite 
correct that this form is not seen in any of the natural languages
that I know.  Of course that isn't saying much.  Rather than saying
descriptively that JCB used this form, I would say that JCB saw 
potentially frequent use (in texts to be written in the future) of
such abstractions, particularly with cleft places, and he developed
an informal procedure to create lujvo for them.  I then reverse-
engineered that procedure.  

There is a set of dikyjvo generated from a directional property plus 
a motion word.  There are fewer words in this set, but it's also easier
(and probably less controversial) to define the combining rule.  Let's
not forget the other flavors.  

> >all I did in cevrirni, if I understand you, is apply a
> >different dikyjvo category than 'normally' applies to final position 'rirni'
> >as if there is a statistical basis yet to make any claims about normality in
> >Lojban lujvo-making.  
> 
> Aargh -we agree! I really should read ahead. Yes, there are no such statistic
> yet, which is why this is not part of dikyjvo as I see them. Jimc would like
> them to be; and they may well become so - but only descriptively and non-
> mandatorily; it's hopeless otherwise.

Key point: maybe something easy can be done so a speaker can indicate
which interpretation he wants for cevrirni.  But before such a mechanism
can be sought, the participants have to agree that dikyjvo are a Good
Idea, i.e. that it's worth searching for the resolution mechanism.  

As to statistics, lujvo can only be counted after the dikyjvo rules are
irrevocably in place, so it's hopeless to base the original rules on 
statistics.  We have to "bootstrap" it just as JCB did:  What will 
people most likely want to say?  How can we make it easy for them?
What place structures and combining rules will coherently generate
large numbers of useful lujvo?

		-- jimc