Re: Rafsi Repair Proposal: 1

[Logical Language Group <lojbab@ACCESS.DIGEX.NET>]

Responding to and:
>> The gismu place structures are also not baselined, but there have been
>> more than enough attempts to rationalize them - the last one took
>> around 6 person-months.
>
>And there is still plenty of room for getting rid of excess places.
>I'd ask Markl to have a go at reforming gismu place structures, since
>those reforms have a chance of getting adopted.

Not much of one, i am afraid.  I can't rule it out completely, but
the selling job would be tremendous.  Every individual gismu place change
has to be tracked throughout the developing dictionary.  I think I have made
a couple of changes since I created te big dictionary fil
e, but the changes
are done manually at at great time cost, so have been generally limited to
indivudal, and specificially to clarificational cases.

>There are still big gaps in the language, though. The semantics is
>only done a little, and the syntax is complete only because total
>garbage is deemed grammatical. By normal standards, there's still
>a lot of design work left to do, and willing designers are surely to
>be welcomed.


For the most part, we intedn that the language semantics will be decided by
actual usage, and not by prescription.  Hence it is not language "designers"
that are needed, but language "users".

>If the discussion of {lo} has retarded the books, the fault does not lie
>with the discussion. First, the discussion of {lo} was not really about
>change, but about what the nature of the present system is. Second, it's
>through discussion that understanding of the language evolves. Third,
>it is futile to write the books under the assumption that the semantics
>is settled. There is no way it can get settled in time for the first
>books, and the textbook presentation will just have to take that into
>account.

One reason I've put the textbook into 3rd place among the books.  It is the
most senstive to semantics changes.

The lo discussion HAS been more serious than you may suspect.  First of a
ll,
over the last year there have been floated perhaps a half-dozen cmavo, all
with undefined syntax, to solve various problems; e.g. the afterthought s
cope
marker.  I am not sure I could even list them all.  Second of all, pending
Cowan's summary proposal he just made on the issue, the discussion of
descriptors in the reference grammar is meaningless without some consensus on
what "lo" means.  Third of all has been that the best minds on the project have
beentied up simply reading (if not responding to) parts or all of the ongoing

extremely dense discussion.  Cowna and I have tried to keep out of it, but
evetually one or tghe othe rof us gets dragged back in as soeone raises a
new issue.  pc has been EXTREMELY helpful in managing the "offcial" position
on the issues, but has been at times himself quite indecisive as to what
HE
believes.  In many ways this issue hits at the logical heart of the language,
and the thought that we might have a deep flaw in that heart has been a nagging
ache that has sapped the nergy of all of the worker-bees.

The discussion of the last month has been delightfully lighter in tone, and
the fact that people have turned back to writing ckafybarja and playing the
phone game rather than arguing deep semantic  issues with potential grmamtical

impact has MUCH reduced the tension here in Fairfax, and has helped get Cowan
back into his noirmal highly productive mode.

>> Note that ANY change to the language that introduces the risk that some
>> speakers will pronounce one syllable as two in word final position
>> completely screws the morphology because you have lost consistency in
>> determining the penultimate syllable to be stressed.
>
>It's not clear that the syllabicity count is determined by the
>pronunciation. More likely it's determined by rule. So there is no
>risk.

There is because this happened before.  JCB had rafsi like "-ria" and "-sia
"
and "rua" that were nominally monosyllables.  But when people said they
 could
not pronounce them as monosyllables, he gave in and said they could do them
as disyllables.  Take the hypothetical lujvo "sainrua" - if rua is a
possiible stresses
are SAI,n,rua  sai,N,rua  SAIN,rua or for a disyllable pronouncer
either sai,n,RU,a or sain,RU,a.  The differences between the mono
syllable
and disyllable pronunciations are so different in stress that the penultimate
stress rule was threatened, and that rule is critical to the unambiguity of
the morphology.  Not to mention that different speakers would pronounce the
same word in unrecognizably different ways.

Not sure what you mean by the syllabicity count determined by pronunciation.
The rules is penultimate stress.  If you NEVER pronounce the optional
disyllables as monosyllables, then following the rules, you will end up
stressing a syllable that doesn't even EXIST in the word for a monosyllable
speaker.

For this reason we totally eliminated optional syallbification, introduced
the apostrophe to force disyllables, and thereby preserve unambiguity.

Oh the other example was one like "ruagra"  RUA,gra (I can't do this one
stressed) vs. ru,A,gra  (which I turn into ru,UA,gra too easily, hence the
need for the devoiced glide that is ' )

>I think it would be less biased to count letters instead, since we
>all agree that {zbasu} has 5 letters, even if we don't agree how many
>syllables or phonemes it has.

But that would change the stress rule, with unknowable effects on the unambi-
guity f the morphology.

>> This is not particularly desired nor desireable.
>
>By me it is.


JCB did taste tests, which while statistically questionable, showed that there

were many instances where even the biased native English speaker base favored
longer polysyllabic words over shorter ones.  The classic example was JCB;s
mother-mother (similar but not the same as ours)  matmatma was strongly
preferred to matmaa.  Indeed, unreduced final terms often were preferred
over monosyllables as well.  Even I at times have used bavlamdjedi instead of
bavlamdei for tomorrow, though I have grown familiar with and generally
prefer the latter.

>> It is ADVANTAGEOUS for lujvo to be longer in syllable counts than
>> gismu (since they are more complex words, and generally also less
>> common words - hence Zipf demands that they be longer),
>
>Some words may get used less than they would were they shorter. To
>judge a word's "natural" frequency, you need to allow for its
>length. To test whether a word is too long, see whether it is more
>frequent than average for words of the same length.

But we don't yet have the kind of frequency data we need for this.  Almost
no lujvo other than the Lojban jargon words have reached a natural-like
level of usage because they aren't known.

>It's a shame, incidentally, that Zipf doesn't apply to cmavo. As
>Jorge has noted, lots of lovely short cmavo never get used because
>they do things like mex and font shifts.


Zipf did apply to the cmavo design.  I could go item by itme through tje
he list and cite several Zipfean reasons for what we did.  You may not agree wi
th the
choices or rationale,m but often the decisoon was made specifically for
Zipfean reasons.

One factor that gets left out of this is that JCB in the 80s went heavily int
o
acronyms in the word set, being used as names, brivla, what have you.
It was
an intractible problem.  And he had half of cmavo space tied up between
upper and lower case Greek and Roman alphabets, all with separate words.
The short shifts WERE our Zipfean solution to keeping acronyms managably
short and sayable, while not tying up half of cmavo space with lerfu.
lerfu have NOT yet reached the level of usage in Lojban that they did in
earlier Loglan, but it is growing in my own usage, and I supsect some of
those issues will run up against our design.

>You should be careful in flinging around this "malglico" label.
>You need to demonstrate that there is a glico bias, and that it
>is (se) mabla.

If it is a bias towards English, it is almost certainly mabla , because the
usage was coined by the dominant English native speaker base of the language.
Basically, we know human tendencies are to make things sound good to our
native dialect ear, and we are therefore unable to unbiasedly judge.

> In this instance, you will need to explain why
>average utterance length does not vary between languages [I don't
>know the exact reference, & haven't read it, but it has been
>claimed]

Well, Harlow has shown wide variance in length of various translations into
and out of Esperanto.  I don't know about utterance length, but Finnish WORDS
are generally longer than English or Lojban ones.  Zipf may not even hold in
some languages.  The Russian word for "use" is roughly ispolzovat' which in
some declensions can be 5 syllables long.  They CAN shorten words, but they
DON'T.  Oh, and I don't know about utterance length, but in written text,
Russian sentences sure SEEM longer in both word count and average word length
than similar English of comparable topic and register.

>But English speakers like them for their etymology and register,
>not for their length. Germans and Russians have ways of shortening
>long words (e.g. gestapo, komsomol, etc).

Yes, but they don't do so except for very common words.  And these are a
bit like the acronyms I just mentioned anyway.

Oh another Russian mouthful - I have always loved this since  it was in
a surprisingly early lesson in my textbook, and the roots are common enough
in Russian text that words of this length are not uncommon:
Pardon my transliteration and probable misspelling
xudozhestvennoi sameldeitlnosti (5+7 syllables) - it means roughly
"amateur art show"

(Cyril is probably turning his stomach at my Russian by now so I'll quit)

lojbab