Newsgroups: sci.lang
Path: cantaloupe.srv.cs.cmu.edu!das-news2.harvard.edu!news2.near.net!news.mathworks.com!solaris.cc.vt.edu!insosf1.infonet.net!internet.spss.com!markrose
From: markrose@spss.com (Mark Rosenfelder)
Subject: Re: Optimal Artificial Languages
Message-ID: <D2o05w.33o@spss.com>
Sender: news@spss.com
Organization: SPSS Inc
References: <3fha85$fal$1@mhadg.production.compuserve.com>
Date: Thu, 19 Jan 1995 18:17:55 GMT
Lines: 135

In article <3fha85$fal$1@mhadg.production.compuserve.com>,
Michael A. Rouse  <73544.625@CompuServe.COM> wrote:
>     I am interested in the componants of an optimal "artificial"
>(constructed) human language -- what would it have in common with present
>languages, and what would be different? How can we maximize the beauty
>(in both aesthetic and mathematical sense) and utility of a language
>without having it either overburdened by detail or too undefined to use?
>Is it possible to chose the "best" componants of language without
>suffering too many tradeoffs in other areas?

>[...]an appropriate analogy, since language as a whole is a method of encoding,
>transmitting, manipulating, compressing, and storing information. 

That's a common view, but misses many of the more interesting uses of 
language, such as its use in stories and poems.  Language isn't just a
matter of wrapping up messages in code; it's also a matter of invoking a
host of implications and associations, some conscious, some not.  See
Sperber and Wilson, _Relevance_.

>2. The language should be the easiest to encode (speak, spell, and sign)
>   and decode (understand). Any increase in complexity in one area should
>   cause a decrease in overall complexity. For example, rather than call
>   something "an underground system where large vehicles ride on metal
>   rails" we can simply come up with a brand new word called "subway."
>   On the other hand, we wouldn't want to give every object a unique
>   name -- it would make a dictionary pretty tedious, for one thing ;-)

The idea that a bunch of "things" exist out there and language just has to
give names to them should IMHO be resisted.  How we divide the world into
categories of things, and how we create new categories, is an interesting
topic; see for instance Lakoff's _Women, Fire, and Dangerous Things_.

>3. The language should be free from ambiguities -- it should be flexible
>   enough to easily add concepts whenever needed, but each word and phrase
>   should have one precise and readily apparent meaning. (I realize that
>   words such as "blue" do not correspond to one specific frequency of
>   light but rather to a range of colors, so it can be argued that it
>   is inherently imprecise; I consider it part of a classification scheme
>   from general to more specific, kind of like "primate" refers to an
>   order including man, apes, and chimpanzees. On the other hand, using
>   "blue" to refer both to color and unhappiness does lend itself to
>   ambiguity -- "The alien was blue" is not very precise, at least
>   without other support words.)

Are you intending your artificial language for human beings or for robots?
Metaphor is basic to how people use language, and extend it to new uses.
If words have fixed meanings and you manage to prohibit metaphor, 
you will also prohibit (or at least inhibit) changes in thought.

>4. The language should be free of unnecessary redundancies in signalling
>   -- no extraneous sounds, letters, or words. Unfortunately, some
>   redundancy might be necessary -- human senses are not 100% efficient,
>   and overlooking a single letter or having a single sound drowned out
>   might change the meaning of something significantly. (Here's a question
>   for any geniuses out there -- is it possible to include an efficient
>   form of error correction in human language? Or is redundancy of some
>   kind necessary to cut down in errors?)

Redundancy *is* a form or error correction.  The information content of 
English text has been estimated at 1 bit per word, which makes it possible
to understand evn f th sgnl s sgnfcnty dgradd.

>6. Another limitation is that no matter what language we come up with,
>   some sounds/letters/signs will be harder to produce than others, and
>   some combinations of sounds/letters/signs will be difficult or
>   impossible to produce. Concepts that are common, important, and
>   interesting should be easy to produce and understand, while those
>   that are impossible, unimportant, and boring should be relatively
>   harder to produce and understand. (I realize this seems to be treading
>   close to the dreaded "Newspeak" of _1984_, but all languages have
>   limitations -- the problem is moving the limitations so they do the
>   least damage.)

And who decides what concepts are important, and what limitations do damage?
You and the Language Control Board?

Natural languages are self-correcting in this regard: if an expression 
comes to seem unwieldly ("omnibus", "taximeter cab", "virtual reality"), 
it gets shortened or abbreviated (bus, taxi, VR).

>7. Finally, the language should be extremely stable, both internally
>   (the grammar which holds the words together) and externally (the words
>   themselves). This doesn't mean it should be inflexible, only that
>   it should not split into incomprehensible dialects just because
>   everyone doesn't live in the same village. This can be accomplished
>   by remembering that *People Are Lazy*. If the simplest word corresponds
>   to the most useful concept, people are less likely to substitute a less
>   meaningful concept. 

Have you any evidence for this assertion?  Linguists generally find that
a particular sound change ultimately affects *all* words that contain its
conditioning environment, not just the "more difficult" words.  Sound 
change can be an extremely stupid process: e.g. in southwestern France,
Latin _gallus_ 'rooster' and _cattus_ 'cat' changed so as to have the same
pronunciation, [gat], an obvious inconvenience.  Again, language users
tend to correct such problems, in this case by substituting a different
word for 'rooster' (which word depended on the region). 

>1. The spoken language is based on a limited number of sounds -- possibly
>   all of those represented by the International Phonetic Alphabet, but
>   perhaps a subset of that (if a sound is hard to produce, easy to
>   misunderstand, or doesn't combine well, adding it might not be
>   worthwhile).

The range of possible human sounds does not form a finite set; rather, 
an infinity of possible sounds exist, but languages divide them up 
(differently) into finite sets.  If an IPA symbol exists, it's because
some human language uses it, so it can't be *inherently* hard to produce--
except to speakers of other langauges.  You might want to read up on
phonology.

>2. The spoken language is completely morphemic -- each phoneme corresponds
>   to a particular morpheme, and each morpheme corresponds to a specific
>   phoneme or combination of phonemes. 

You've been reading too much Heinlein.  I doubt very much that a language
with several hundred phonemes would be practical; and even several hundred
morphemes wouldn't suffice for a respectable language.  Eliminating the
redundancy of natural language is also an invitation to miscommunication.
Imagine if misconstruing a single phoneme didn't simply substitute an
erroneous retter, but a whole different banana-peel.

>3. Sounds are combined into 1-syllable words covering the breadth of
>   human knowledge -- well, if it's good enough for the Chinese, it's
>   good enough for me! 

If Scott Horne reads this far, you're in trouble...

>8. Sign language should follow spoken and written language. Finger-
>   spelling (or the equivalent) should be based on the phonetic/morphemic
>   forms of spoken and written language. 

Deaf people won't like this.  They find close equivalents of spoken
languages, such as Signed English, much more difficult and annoying
than natural sign languages, such as ASL.
