Hallo!

On Mon, 13 Sep 2010 20:51:43 +0100, R A Brown wrote:

>  On 13/09/2010 13:02, Jim Henry wrote:
>
>  [...]
>  >  Even within the Huffman coding solution space, it would
>  >  be trivial to come up with a solution that's equally easy
>  >  to code as Plan B, but significantly easier for humans to
>  >  parse
>
>  Yes, indeed - I've done it   ;)
>
>  >  (though still probably nowhere near as easy to
>  >  parse as those that involve fixed-length morphemes, or
>  >  subsets of initial and terminal phonemes or syllables).
>  >  For instance, plosives occur at the beginning of
>  >  one-syllable morphemes, fricatives at the beginning of
>  >  two-syllable morphemes, etc.
>
>  Agreed - I don't a Huffman coding solution (i.e. initial
>  sound/syllable determines morpheme length) is ever going
>  feel 'natural' to humans.

Indeed not.  But most self-segregation schemes won't feel
natural to humans (to what else?  Aliens are not yet known;
animals don't use full-fledged languages; computers do not
have feeelings).  Some may be so subtle that they don't feel
unnatural - but in those cases, people will be likely to
entirely miss the fact that the language is self-segregating
at all!

>  >  Something that we haven't discussed here recently,
>  >  though, is *how and when* a self-segregating morphology
>  >  is or would be helpful.  Of course it would make text in
>  >  an engelang easier to parse with software than that of
>  >  any natlang, or even fairly regular conlangs without SSM
>  >  such as Esperanto.
>
>  I'm told in fact, that Esperanto is not particular easy
>  language to write a parser for.  But I admit I've not tried it.

Certainly, SSM makes it much easier to parse the language.
At least, you can clearly tell where morphemes begin and end.

>  As far as self-segregating morphemes are concerned, it
>  depends upon the language. I know when I was learning
>  Speedwords in the late 1950s I found it quite confusing not
>  being able to spot morpheme boundaries. I think in
>  briefscripts and certain types of engelang it is helpful for
>  _humans_ to know unambiguously where morpheme boundaries occur.

Yes.  In natural languages, the large amount of redundancy makes
it easier to understand the language even without self-segregation.
Whenever you miss a morpheme boundary, you'll soon notice that
the way you try to parse the utterance it becomes ungrammatical,
and retry.  We all do that subconsciously, and therefore need
no self-segregation in our native languages.  But in a dense
language such as Speedwords, which squeezes out just about any
redundancy it could, I can easily understand that self-segregation
is a good thing.

>  >  But, as Logan pointed out upthread,
>  >  humans are really good at pattern-recognition, and
>  >  usually identify the morpheme boundaries in languages
>  >  they're familiar with without problems, even in languages
>  >  that present a lot of ambiguities to a typical software
>  >  parser.
>
>  Yes, agreed. Certainly if one is aiming to produce a
>  naturalistic looking artlang self-segregating morphemes are
>  not wanted. As I said above, it depends on the language.

Sure.  In a naturalistic artlang (such as the ethnic language
of a fictional human nation), self-segregation is not called for.
That is, unless one finds a way to arrive at self-segregating
morphology by a set of natural-looking phonological and grammatical
changes.  My conlang Arne at least gets close to self-segregation
(basically the Konya way), as I have observed further upthread.
I think I could nudge it to true self-segregation without breaking
its naturalness; now that's a challenge.

But the "natural" habitat of self-segregating morphology are of
course engelangs of all stripes.

>  >  When and how and how much would SSM make things
>  >  easier for humans?
>
>  When learning Speedwords    ;)

Yes, when learning a foreign language, you search for morpheme
boundaries a lot, and a self-segregation rule makes that easier.
Hence, it would not be a bad idea to have a simple self-segregation
rule in an IAL.

>  [...]
>
>  I can well imagine that in Volapük, where words can grow
>  quite long, it would be useful know unambiguously where
>  morpheme boundaries occur.

Yes.  Volapük is an inflection-heavy monster.  Perfectly regular,
yes, but baroque in its morphological richness.  And *ugly*.
(Not that I dislike richly inflected languages; but Volapük
does it badly.)

>  [snip]
>  >
>  >  On the other hand, for a real-time spoken-language
>  >  situation -- a conversation or lecture, for instance --
>  >  I'm not sure any SSM scheme more complex than
>  >  fixed-length morphemes would give a significant edge over
>  >  general human pattern-recognition abilities.  One would
>  >  have to experiment to find out.
>
>  You may well be correct. Piashi (which I've now abandoned)
>  was going to have fixed-length morphemes.

So does X-3 - one phoneme, one morpheme.  The only exception
being proper names.

--
... brought to you by the Weeping Elf
http://www.joerg-rhiemeier.de/Conlang/index.html