TECH: ungrammatical Lojban

["(Logical Language Group)" <lojbab@GREBYN.COM>]

It is NOT "official policy" that ungrammatical Lojban is 'not Lojban', as
suggested by And Rosta.  I'm sure that there is no policy on the matter
whatsoever.

I personally feel that text within "zoi" quotes is by definition not Lojban,
and that text with "lo'u/le'u" quotes is by definition to be assumed
ungrammatical and hence unparsable.

If text is not explicitly marked as non-Lojban or non-parsable by those quotes,
I as a Lojban reader or listener will attempt to parse that text.  Thus
said, I will note that Lojban has a problem with ungrammatical text error
recover in that the language is designed to allow the speaking of grammatical
nonsense.  THus, when you make a grammatical error such that the result does
not parse, the listener often has an extra burden in trying to establish
your intent: were you trying to express the most plausible error-corrected
version, or not?  Ando fo course, with no fluent Lojban speakers or listeners,
the question of what is "most plausible" is totally up in the air.

My own experience as one of the more skilled Lojban reader/listeners is that
I have a very low tolerance for most kinds of grammatical errors at normal
communication speed.  If someone leaves out a "cu", but pauses before the s
selbri, I will usually figure it out without a delay (and such pauses happen
a lot due to phrasing syntagmas).  If someone makes an error such that it
apeears that there is a second selbri starting when I have already grasped
at a n earlier one, then I have to wait for the whole sentence to finish,
run the sentence through my head a couple of times, try to figure out what
error MIGHT have been made, then proceed on the basis of asssuming that error
correction.  A sentence with more than one grammatical error is thus usually
hopeless, or at least I won't figure it out fast enough to converse.  Even
a single error often makes me unable to keep up with current non-fluent
speech; likewise a momentary failure on my part to recognize an unfamiliar
gismu or lujvo (which is my failing and not the speakers or the grammars
problem).  Thus I had big problems with Nick's surprise call last week, because
he not only made a couple of gramatical errors that he himself was later
able to identify, but because he sylistically differs from the people who
are talking around here, he was using some gismu that have been rarely heard
in consversation around here.  Thence he had to repeat himself several times
before I grasped just what word he was saying.

An especially bad grammar correction problem for me is an unmarked "lenu"
clause. "*mi klama le zarci cu xamgu" would almost certainly be totally
unrecognizable to me in speech without several playbacks.  The "obvious"
to me as concocter of the example, corr4ection is to add "lenu" at the
beginning.  But I can see at least 3 or for other equally minor cvhanges that
give a grammatical Lojban sentence (including adding a "lenu" before 'le xarci'
and changing the "mi" to "me" or "mo" or "mu" or "ni", all of which make for
some nice grammatical nonsense).

The capability to meaningfully do Lojban grammar error correction will have to
wait until we have fluent enough Lojban speakers that we can identify what
a typical error made by a fluent speaker will be, which may be quite unlike the
hodge-podge of erros made by learners even up to my level of skill.  As such,
the standard for "good Lojban" will have to remain that it passes the parser
as grammatical and is broken down by said parser to group in the way the
speaker intended.  But I won't call erroneous Lojban "not Lojban", just
"erroneous Lojban", priovided that there is some capability to extract
communicative meaning fromit.  (Some of Michael Helsem's poetry alas stretches
this boundary way too far.  But then Nick seems to be able to figure amounts
of Michael's stuff out that I cannot, so the boundary is fuzzy on this issue.)

lojbab