Hallo!

On Wed, 22 Jul 2009 17:13:51 -0400, Jim Henry wrote:

> On Wed, Jul 22, 2009 at 11:37 AM, Bruno
> Barcelos<[log in to unmask]> wrote:
> > What do you think of a language which made its vocabulary short using
> > letters like ý (y with an acute) or Å (a with a ring)
> > Using those we would have more combinations of words with 2 or 3 letters,
> > and also words like "zzn", even being unpronounceable, would be accepted,
> > since the language would be written only. That could be considered a
> > linguistic shorthand, which can be written fastly...
> 
> It sounds like you are talking about two different things: compactness
> of the written form, and speed of writing.

Indeed, they are two different things.  And compactness of the spoken
form, admittedly a non-issue in Bruno's case, is yet another.

>       I'm not sure they can both
> be maximized at the same time.  Maximum compactness suggests, as you
> say, using lots of non-ASCII characters; maximum speed of writing
> suggests using only characters that can be typed easily with a
> standard keyboard.  For either criterion, either making the language
> written-only, as you suggest, or giving it a large phoneme inventory
> and allowing many possible combinations of phonemes, would tend to
> increase the number of possible word-forms and thus conciseness.

If the language is written-only, you need not worry about such things
as phonotactics; you can use any sequence of characters that may
occur.  This is an issue, however, in a language that is also meant
to be *spoken*, as in Heinlein's Speedtalk, where phonotactics opens
a back door through which redundancy might creep in.

> You also need to look at how you assign words to meanings: meanings
> that will be referred to frequently should get assigned the shortest
> words, while meanings that are referred to less often should get the
> longer word-forms.

Most natlangs at least get close to this (isn't this observation called
"Zipf's Law"?); new concepts often receive long words, but if these new
concepts become commonplace, people tend to replace the frequently used
long words by shorter ones, either by shortening the long word, or by
extending the meaning of an already existing shorter word to cover the
new concept.

In a true speedtalk, however, *all* morphemes are as short as they can
be - just one phoneme.

>       Ideally you would do this empirically, with 
> actual frequency data from analyzing a corpus.  But frequency data
> will vary from one language to another; frequency data from various
> natlangs and conlangs won't reliably tell you how often concepts will
> be referred to in your conlang, with its unique grammatical and
> semantic structures.  You could use a build-one-to-throw-away
> methodology, where you start out with deciding on your  grammar and
> semantics, and make a temporary phonology and set of word-forms; then
> write text in the conlang, then analyze the corpus and relex it with
> shorter forms for the more frequent words and longer forms for the
> less frequent words.  (I'm doing this with my own säb zjed'a (
> http://www.pobox.com/~jimhenry/conlang/conlang13/intro.htm ), but I'm
> not aiming purely at maximum conciseness, but at a good balance
> between conciseness, euphony and noise-resistance.)

An interesting experiment!  This is probably a better way to build a
maximally efficient language than the brute-force attempt of Speedtalk
with its monophonemic morphemes.

> You could take a look at the recent discussions in the list archives
> about written-only languages; Brett Williams is working on one based
> on the QWERTY keyboard.  Also, there are a number of conlangs designed
> to be concise.  See John Quijada's Ithkuil,
> 
> http://www.ithkuil.net
> 
> Dutton Speedwords (mentioned by other posters in this thread), and Ray
> Brown's Piashi:
> 
> http://www.carolandray.plus.com/Briefscript/Index.html
> 
> R. Srikanth's Lin:
> 
> http://www.iiap.res.in/personnel/srik/Lin.html

I also have an experimental highly concise language in the making
which is named "Quetech", from the native expression /k_w@t@C/
'language of small time', i.e. 'speedtalk'.  Quetech is an
experimental language meant to test how well a speedtalk would
actually work, especially how much it actually shortens utterances
as compared with "ordinary" languages.  There are two reasons why
it might fall short of its goal:

1. Phonotactics - in a spoken speedtalk, you need to restrict the
   possible sequences of phonemes in order to get pronuncible
   utterances.  This opens a back door through which redundancy
   may creep into the language.

2. Oligosynthesis - if all morphemes are monophonemic, the number
   of morphemes is limited by the number of phonemes, which cannot
   practicably be increased above a few hundred.  You will need
   compounds to express many concepts expressed by roots in "normal"
   languages.  Reality is too complex to pigeonhole it into a few
   hundred categories.

How do I address these problems in Quetech?

1. All native Quetech morphemes are exactly one phoneme long (as in
   Heinlein's idea).  Lexical morphemes (including pronouns) are
   consonants; these serve as nouns, verbs, adjectives and adverbs.
   Grammatical morphemes, such as prepositions and conjunctions,
   are vowels.  Each clause begins with the predicate (a lexeme)
   followed by one or more arguments, each being a lexeme predeced
   by a grammeme encoding its semantic role.  You just get a sequence
   of alternating consonants and vowels, beginning and ending with a
   consonant.  Attributes are expressed by lexemes following the head
   of the NP, to which they are connected by a grammeme.

2. In order to get around the oligosynthesis problem, Quetech provides
   an escape mechanism allowing for the importation of arbitrary
   lexical material.  Borrowed lexemes can be of any length and are
   enclosed in glottal stops; the glottal stop does not occur anywhere
   else in the language and may not occur within borrowed lexemes.
   Thus, self-segregation is preserved.

When I have developed the language sufficiently, I am going to test
it by translating a text corpus into it and counting the phonemes and
syllables, comparing them against English and German.  I will perhaps
also ask John Quijada to translate my corpus into Ithkuil, Jim Henry
to translate it into säb zjed'a and Ray Brown to translate it into
Piashi, to get additional results for other highly concise conlangs.

... brought to you by the Weeping Elf