Hallo!

On Thu, 19 Jul 2007 18:29:52 +0100, And Rosta wrote:

> John Crowe, On 17/07/2007 14:32:
> > Efficiency (use less space/time to convey the same amount of info) in
> > conlangs is one of my newer interests, and I don't know much about this
> > idea. What is the formal name for this? 
> 
> In engelinguistics we call it 'concision'.

And note that redundancy (the opposite of concision) is not a bug
in human langauges, it is of course a feature.  But that should not
stop anybody from designing maximally concise engelangs.
 
> > What are the results of formal studies/works in this area? 
> 
> There aren't any, really. The closest thing would be compression
> algorithms such as Huffman encoding. 

Jim Henry proposed this last year (April 18, 2006):

http://listserv.brown.edu/archives/cgi-bin/wa?A2=ind0604C&L=CONLANG&P=R3801
(original post)

http://bellsouthpwp.net/j/i/jimhenry1973/conlang/conlang13/intro.htm
(on his website)

The basic idea, as far as I understand it, is to build a language,
translate texts into it, measure the token frequencies of the
morphemes, and then relex is using the shortest morphs for the
most frequent morphemes.

> > As far as I know, only a few conlangs strive to be 'efficient'.
> > A few of the
> > following are from Richard Kennaway's much linked to but outdated list,
> > which is still useful because it has descriptions on the list, making it
> > possible to search for specific things in conlangs.
> > 
> > *Speedtalk [Heinlein]. of course, comes to mind.
> > 
> > *Ithkuil (and Ilaksh) [Quijada] Relatively well known, if I'm not 
> > mistaken. 
> > Logical/Philosophical
> > 
> > *Lin (or Ln) [Skrintha] A bunch of broken links. No attempt to be
> > logical/philosophical.
> > 
> > *Mindbrush [Gressett] "allow faster, more efficient thinking"
> > 
> > *Earth Minimal "An Ultra-Compact Auxiliary Language" Seems to focus on
> > things other than efficiency more.
> 
> IIRC, Speedtalk's concision was due to a large phoneme inventory. 

Yes.  The idea was:

1. Use as many phonemes as possible (throw in every distinction found
   in any language);

2. Use as few morphemes as possible (as in Basic English);

3. Increase the number of phonemes and decrease the number of
   morphemes until both numbers meet in the middle;

4. Now you can have each morpheme exactly 1 phoneme long.

This is an idea worth exploring; I am planning to do that with
X-3, but I have put aside that project as I am more interested
in other things (such as exploring the Albic languages, which
are just as "inefficient" as most natlangs), and closed-vocabulary
schemes are far from being without problems: most things would
have to be cicrumscribed - but if all morphemes are just one
phoneme long, a six-morpheme word would still be just six
phonemes long, i. e. not very long at all.  Another problem is
of course the phonotactics - how do you avoid ending up with
monstrous consonant clusters?  You need something smart;
in X-3, I am experimenting with a scheme in which nouns are
consonants, and verbs and adjectives are vowels.

Ray Brown, in his Briefscript project, goes a radically different
road - one can say that he sets off in the opposite direction
in order to reach the same goal: he uses so *few* phonemes that
he can use the Latin alphabet as a syllabary.  His language
may not be highly concise in spoken form, but in the written
form, it becomes quite concise because any letter combination
is pronounceable and may have a meaning, while in natlangs,
most letter combinations (e.g. "xbrlynpha" in English) are
meaningless noise.

> Ithkuil's central technique is, it seems to me, a kind of semantic
> template that is both rigid and rich. The rigidity makes it possible
> to assign to it a maximally concise encoding. The richness makes it
> sufficient for speakers' communicative needs. 

It helps that the language has so many phonemes.

> Concision was Lin's sole aim. One of its original techniques was
> systematic largescale homonymy without ambiguity -- which allows
> the same short forms to encode many different meanings.
> 
> I don't know owt about Mindbrush or Earth Minimal.

Never heard of them, either.

> > I have read some essays concerning half-related topics, it seems that a 
lot
> > of people that human language cannot be made more efficient. "Amiguity is
> > necessary." 
> 
> Not clear what you mean. I reckon the design of natural language
> can be improved upon -- e.g. made less ambiguous without being
> less concise.

This is something I am mildly doubtful about; sure, no natural
language is perfect, but trying to optimize a language for a
particular parameter probably leads to languages with other
shortcomings.

>       But speech communities collectively evolve languages
> that meet their needs; so natlangs are all 'good enough', and speech
> communities will stick with their current language so long as it
> is good enough.

Yes.

> > Even after setting aside human read/parseability, it still seems hard to
> > draw the line. Theoretically, if a language has maximum efficiency, then
> > there must be no such thing as an incomplete utterance or an ungrammatical
> > statement, i.e. every possible utterance (or combination of symbols) 
> > within 
> > the rules means something. 
> 
> Yes, every string of symbols would be meaningful, but it wouldn't
> necessarily be a complete sentence. 
> 
> Note that 'ungrammatical statement' is ambiguous. If it means 'string
> of symbols without meaning', then your maximally efficient lang would
> indeed lack ungrammatical statements. If it means 'pairing of form and
> meaning that is inconsistent with the form--meaning correspondence rules',
> then of course there would be ungrammatical statements aplenty.

Well, you can hardly have a language without grammar; so you won't
get a free monoid over your alphabet in which every possible string
is meaningful.

... brought to you by the Weeping Elf