Re: Impossible Gibberish (was Re: On the design of an ideal language)

On 5/2/06, Sai Emrys <sai@...> wrote:

> FWIW, what I meant by the PSC is not necessarily that every utterance
> must make sense in some *pragmatic* sense (vis. colorless green ideas)
> - that's impossible AFAICT - but that it must be *parsable*.
>
> E.g. "I eat five apple yesterday." That's just "wrong", and in fact
> it's wrong in a way that could only be reasonably interpreted as one
> thing - "I ate five apples yesterday". That one is "correct" and the
> other not is a waste of semantic space. I would have the two sentences
> mean different, but both parsable (and plausible in that sense).
This particular kind of ungrammaticality is easy enough
to avoid in a conlang; make the plural marker optional
when another number word is present, and the tense
markers optional when a specific temporal complement
(like "yesterday") is present.  Nonsense like "I eat
future yesterday apples" would still be possible, where
you have a tense marker that contradicts the temporal
complement; but I reckon that's the sort of
"colorless green ideas" that you aren't trying to rule out.

>and second, where
> such relatively simple lexemes are NOT real words, but really long
> complex ones are.
If you use up all or almost all of the monosyllabic
word-shapes your language allows before
creating any disyllables, and so forth, then the
language will have no noise resistance; the
slightest amount of background noise or the
most trivial typo will make a sentence mean
something different and in many cases
equally plausible in context.

> FWIW, the PSC and the (implied, but should be added explicitly)
> Principle of Noise Resistance (per previous poster) are of course at
> odds to a certain extent.
>
> I would argue, though, that if you were going to build in redundance
> for the sake of a PNR, it could take on a much more elegant/efficient
> form than this sort of rule.
If you mean "no two morphemes differ by fewer than
two phonemes", I agree.  This is an experimental
language and the redundancy criterion will probably change in a future
revision.  If you think "no two morphemes
differ by fewer than two distinctive features" is still
too inelegant or inefficient, what do you suggest
as an improvement on that?

Actually, I think "minimum two different distinctive
features" is maybe not noise-resistant *enough*, and
"minimum two different phonemes" wastes space.
A good compromise might be "minimum three
different distinctive features", but I'll need to
make extensive revisions to my vocabulary generation
script to handle that efficiently.


> P.P.S. I would like to add that I would like to find a niche for
> Jabberwocky-style not-quite-nonsense... but it does conflict with this
> pretty directly. Except, of course, if you carve out some way to have
> the sort of onomatopaeic / associational semantics that are implicit
> in Jabberwocky. It does make one think of things, after all. There
If you define a redundancy criterion that
suits you -- for instance, no two morphemes
differ by fewer than two distinctive features
-- and then use up as many monosyllabic
words as you can consistent with that redundancy
criterion, then any mutation of one distinctive
feature in one of your words would produce
a nonsense word that's halfway between two
real words (it sounds equally similar to both).
Maybe that would allow Carrollesque poetry.
For instance, if "ba" and "mi" are real words
then "bi" or "ma" would be nonsense-portmanteaux
that suggest "ba" and "mi" simultaneously.

....I also note that neither Sai nor And includes
self-segregating morphology in their criteria
for an ideal language.   In And's case, he
has something else that's just as good in
disambiguating a parse string for a fluent
speaker, though not as helpful for a learner
as self-segregating morphology would be.
In Sai's case I suppose that self-segregating
morphology would be too great a constraint
on filling up the phonological space with
real words, perhaps?

My engelang's phase 1 has self-segregating
morphology, and probably the next phase or
two will as well, but later on I may go with
something less restrictive that's roughly
equivalent to And's Livagian rule.

--
Jim Henry
http://www.pobox.com/~jimhenry

Reply