Auto-Segmenting Engelang (was REQUEST: Engelang?)
|From:||Jörg Rhiemeier <joerg_rhiemeier@...>|
|Date:||Tuesday, May 31, 2005, 16:59|
On Tue, 31 May 2005 02:57:57 -0400,
Rob Haden <magwich78@...> wrote:
> One such problem that I frequently grapple with is auto-segmentation. That
> is, can a language be constructed such that any combination of morphemes
> has one and only one meaning?
Yes. First of all, assign only one meaning to each morpheme.
Then, design the morphemes such that your morphology is self-
segregating, see below.
> This problem is also very interesting from a
> natural-language-processing or speech-recognition point of view. Any
> takers here? :P
Actually, this is quite easy to achieve, and many engelangs do that.
It is called "self-segregating morphology", and it is achieved by
designing the morpheme structure constraints such that morpheme
boundaries are marked by phoneme combinations that could not occur
within a morpheme.
A simple example of self-segregating morphology uses the following
1. Each morpheme begins and ends with a consonant.
2. No consonant clusters occur within morphemes.
These rules limit the morpheme structures to C, CVC, CVCVC, etc.;
whereever a consonant follows another consonant, must therefore be
a morpheme boundary between the two consonants, and nowhere else.
AFAIK, self-segregating morphology is either non-existent or at least
exceedingly rare in natlangs, probably because sound change would
soon foul it up, and should thus not be used in naturalist artlangs.
In engelangs, however, it is just fine, and it definitely increases
computer tractablity etc.