Hi, Gary, Larry and mo',
Let's begin at the beginning:
"The Kalusa experiment ( http://kalusa.fiziwig.com/index.php ) was an
attempt to
> build a conlang collaboratively by allowing anyone to contribute any
random
> words and sentences, and to have other participants vote on the
suitability of
> those contributions."
These were pretty much the parameters for Kalusa, weren't they?
It seems a tad unrealistic to now turn about and complain that this
"open-ended" experiment didn't turn out the way we would have
liked it to!
Allow me to draw a parallel. For most of my working life, I have
been designing and constructing software systems. Although
implementations change with new technical resources, some basics
of systems design were established in the early days of general
systems theory, and these basic principles reflect some immutable
needs. One of the common models for systems development uses
the notion of "the systems life-cycle". An entire system's life is
regarded as comprising several distinct phases, each conducted
more or less formally or effectively. Those phases are:
1. Analysis
2. Design
3. Development
4. Testing
5. Implementation
6. Review
Effective systems spend most of their life being used, in the
Implementation phase. For that to happen, it is essential that
the Analysis phase first discloses the needs that system must
meet; the Design phase next invents ways to meet those needs;
the Development phase creates the means to implement those
ways; and the Testing phase ensures that the developed system
meets the original needs. Finally, after the system has been in
use for a while, it is reviewed to determine whether the needs
have changed and whether the system still meets the needs.
The foundation-stone of the whole system is correct analysis.
Failure to discover a need means that the system designed and
implemented can only ever meet that need by chance - by sheer
good luck.
Against this background, it seems to me that perhaps the
essential cause of your dissatisfaction with the outcomes of
the Kalusa experiment is an incomplete analysis of your
requirements. However, the good news is that we can learn
from this experience and incorporate new design goals for
Kalusa Mark 2.
Also, when I last looked at Kalusa - admittedly some months
back - it struck me as phenomenally successful in meeting its
stated goals. Minor dissatisfaction with one aspect of the
achievement should be seen in perspective, and ways designed
to avoid a repetition.
Possibly the most effective way to rescue Kalusa from its
unfortunate detour would be - if possible - to backtrack it
several weeks (I hope you've had regular backups?); put new
constraints in place to reduce the chance of repeating the
undesirable behaviour; and resume running, monitoring the
results at least once a week for suitability.
I don't know the actual means by which Kalusa lost its way-
Was it hijacked? Weren't enough people paying attention?
Were a few users very active, with new words and structures
they proposed by disproportionately used and approved by
themselves? An examination of the way it went wrong may
suggest suitable corrective measures; for example:
1. Should I be allowed to vote on my own proposal?
2. Should any new word be accepted into the corpus with
fewer than 5 votes?
It'd be a pity to see the enthusiasm and effort that's gone
into Kalusa completely wasted due to avoidable design faults.
Unfortunately, Gary, the actual history of Kalusa reflects
the importance of your ideas 1) cultural context and 2) basic
needs; by showing that those "needing" to establish "Seussian"
sounds can have their way if not actively and vigorously
countered. I agree those notions are important, but the
hardest to establish objectively may be the "basic needs".
It's therefore in this area that most work needs to be done
(Analysis) BEFORE ever the Design and Implementation are
attempted.
BTW, I note that your examples of basic needs, while well-
suited to, say, an IE people at any time in the last five millennia,
would be completely inappropriate to any Australian culture at
any time in the last 40 millennia (excepting the last two
centuries in both cases). That doesn't matter, however; what
does matter is that the experiment begins with very clear
notions of the needs the language must meet.
Also BTW, the sound of many Austronesian languages may
strike you, as many other non-native learners, as rather
"babyish", simply because they make extensive use of
reduplication to derive words from more basic ones. But if you
don't want the language to use those means or create those
sounds, you need to define this as a need up front, and accept
no design as adequate that will permit them.
I don't actually like your idea 3) canonised vocabulary, since
it could easily play into the hands of would-be hijackers. All
they'd need to do would be take the Basic English vocab, for
example, and generate a set of unwieldy and ugly gibberish
words for them (using, eg, the magnificent "wordgen"), get
them voted on several times. Providing they were quick enough,
this would effectively sabotage any chance of success in the
resulting collaboration.
No, Larry, having to create an interlinear would put off many
contributors with suitably simple and practical ideas for new
vocabulary and constructions. I don't think that would help.
But requiring people proposing a new word to explicitly define
it doesn't seem too onerous. Most words, particularly nouns,
will correspond one-t-one to existing words in some natlang;
a new kind of construction needs to be exemplified (and possibly
counter-exemplified) in order to delimit its range of application.
As you wrote: "there should be enough information that folks
can see what is meant by each morpheme and each word."
We need to think of ways to enforce this; eg, a sentence
containing a new noun must have a compulsory pair of fields to
give an English equivalent for the noun in the context of this
sentence; any sentence using a noun should have a similar pair of
fields by which the user can indicate an extension of the noun's
meaning in this context; any new construction should require at
least three example sentences, perhaps including a negative
instance if applicable. Proposed sentences could be flagged with
a list of warnings of potential rule violations eg "this construction
has no negative example" for the information of voters.
Having said (all!) this, I invite you to open up discussion on the
needs of your ideal collaborative language, beginning by posting
all those needs of which you are already fairly certain.
Now I'm off to have a geek (a gander, a captain cook, a look) at
Kalusa, to see how closely it approximates any natlang of my
experience to date.
Best wishes for happy experimenting!
Yahya
---
On Sun, 27 Aug 2006, Gary Shannon wrote:
attempt to
> build a conlang collaboratively by allowing anyone to contribute any
random
> words and sentences, and to have other participants vote on the
suitability of
> those contributions.
>
> Being entirely open-ended meant that people could contribute whatever
words and
> sentences they wanted to. Unlike a real pidgin or contact language, or
language
> developed within a community of people sharing work and social activities,
> there was no pressing need to develop the means to say certain "necessary"
> things like "I need two pounds of rice and a cabbage." or "Help me unload
this
> crate of nails." or "Come quickly! The goat has fallen into the well."
>
> As a consequence, a great number of "frivolous" sentences and frankly
> ridiculous words have been contributed. How often will we need the word
for
> "hyperinfracaniphilia", or "epistemic"?
>
> Not being a spoken language, apparently little attention is being paid to
the
> sound of the language, and words and sentences that are either unpleasant
> tongue twisters, or frankly childish sing-song constructions have found
their
> way into the language. How many reduplicated words does any one language
need?
> Sentences that begin to sound like "Hong Kong King Kong sing song ping
pong
> ding dong." are not the kind of things one would hear in a real spoken
> language, and yet such grotesque words are proliferating: "Zotasota
feniseni
> rofkosofko onasona irusiru ishisishi zokusoku fezosezo." making the
language
> begin to sound like something created by Dr. Seuss while on mind-altering
drugs.
>
> A side effect of flooding the vocabulary with such words is that the web
page
> that shows the most recent contributions is so saturated with these
"goofy" and
> useless words that many of the participants and contributors become
discouraged
> and leave the project because the "real" sentences have been buried under
a
> deep pile of ickysicky kakasaka dudupudu, which if not actualy destructive
to
> the language, is certainly not productive of a usable language.
>
> Comments from users include ones like this: "Ack! I go away for three
weeks,
> and return to find Kalusa defeated! Oh well. It had a good run.
...[addressing
> certain contributors]... Looks like you've successfully driven everyone
off,
> including me."
>
> While there is a lot good stuff in Kalusa, and I fully intend to continue
to
> keep the website going, I can't help but think that what I've learned from
> Kalusa could be applied to a much better collaborative conlang project, so
I
> have to ask what was learned from the project, and how can these problems
be
> prevented in any future collaborative language project? Here are some of
my
> ideas to help keep the project more focused.
>
> 1) Language and culture probably evolved together, and the cultural
context
> would be important to the development of the language. Therefore it would
be
> helpful to provide at least a basic cultural context for the language.
>
> 2) The earliest utterances of the language should deal with the most basic
> daily needs of the people who speak the language, and not with
"existentialism"
> and "hyperinfracaniphilia". Therefore, rather than allowing contributors
to add
> random (and often "goofy") sentences and words, a large collection of
simple
> sentences dealing with the daily concerns of the people would be provided
in
> English; sentences such as "It is time to plant the corn." and "Father has
gone
> to the marketplace.". Contributors would suggest translations for the
sentences
> in this corpus of daily life, and all of the different suggested
translations
> would be presented together on the web page where they could be voted on.
Once
> a clear winner emerged the sentences with lower vote rankings would be
deleted
> and only one "correct" way to translate that sentence would be retained.
Minor
> variations in emphasis or shades of meaning might be retained, but
translations
> that departed radically from the highest ranking translation would be
discarded.
>
> 3) Languages do change and evolve, but they also exhibit a great deal of
> stability in their most basic vocabulary. Basic vocabulary would be
stabilized
> as soon as a word emerged as clearly the most popular translation, and
these
> "standard" words would be added to a cannonical dictionary. Translations
that
> radically violated this cannonical basic vocabulary would be dropped,
since,
> for example, once the word "kaya" had been well established for "water" it
is
> unlikely that the word "gumisumi" would suddenly take its place. Anyone
who
> suddenly began translating "water" as "gumisumi" would clearly be
considered as
> being in a state of sin, and those translations would be expunged from the
corpus.
>
> What other suggestions does anyone have for creating an improved
collaborative
> conlang project?
>
> ------------------------------
> Larry Sulky replied:
>
> Gary, as always your posting is incisive and thoughtful. I think your
> ideas for an improved go-round are spot on. One additional
> consideration might be this: more clarity as to which morpheme in
> Kalusa2 corresponds to which morpheme in English. I'm not saying you
> should require folks to submit an interlinear (I'm kind of thinking
> it, but I'm not saying it), but there should be enough information
> that folks can see what is meant by each morpheme and each word.
> Besides, that would be a necessary condition in order to establish
> that, for example, "kaya" is the word for 'water' and thereby be able
> to canonise it.
>
> ------------------------------
--
No virus found in this outgoing message.
Checked by AVG Free Edition.
Version: 7.1.405 / Virus Database: 268.11.6/428 - Release Date: 25/8/06