Re: Frame-based vs ontology-based vocabulary

I'm not sure what to quote here, but I feel like I should quote
something, so here it is:

Sai wrote:
<<
So instead of saying something like "I bought a game from the market"
it might be something like "[commercial transaction](buyer = I, place
= generic, money = ?, object = [game](instrument))". Except of course
much more sexy than that; that's just the analytical version.
 >>

Ever heard of a language I did called Megdevi?

http://dedalvs.free.fr/megdevi.html

It was my first one.  It had a lot of problems, which you've heard
some about (and more are detailed on that page), but as for what
was good about it, go to this section:

http://dedalvs.free.fr/megdevi.html#na

And take a look at the numbered lists (there's one made numbered
list in this section that has paragraphs and unordered lists
intervening;
the numbered list is relevant here).  What you described in your
e-mail is what I, essentially, tried to do with Megdevi, and it was
directly inspired by Arabic.

Arabic, obviously, is a natural language, and has been subject to
historical change--the "elements", as it were--but the way its
vocabulary is set up mirrors in some ways what you're talking
about.  For example, if you take a small subset of the vocabulary,
you can see that you have, for example:

waalid = father
d_?\aabit = soldier (last /t/ may be pharyngealized; I forget)
t_?\aalib = student

I can't remember much Arabic anymore, but what you have here
are a bunch of words with the form 1aa2i3, where the numbers
are the consonants in the triconsonantal root.  If you look at a
subset of the vocabulary of Arabic, you can form the generalization
that the pattern 1aa2i3 is associated with animate agents (default
male).  You can then look at the roots and see that they all have
to do with a particular category (so /w-l-d/ is something like the
family.  The word for "father", "mother", "boy", "children" are all
in there; /d_?\-b-t/ probably has something to do with fighting;
/t_?\-l-b/ has to do with study).  An actual word, then, is the
intersection of the two.  Over time, individual words come to
refer to things ontologically, but the idea is that that's not the way
it started out.

I saw this system from Arabic, and decided to try to make a regular
variant for Megdevi.  What I just showed you was my attempt.
I found it difficult, of course, and rather than revising my work,
I ended up creating a bunch of affixes, but if you view that as a
starting point, you can create a language that could work.

Using your terminology, then, the frames would be separated
into two parts: specific frames (this would be the triconsonantal
roots of Arabic and Megdevi), and agent roles (this would be
the vowel patterns).  The problem with Megdevi is the instantiation--
not the idea itself.  For example, for your purposes, the "natural
noun" would be far too vague, in that it covers animate agents
of actions ("teacher" from the "teaching" frame) and roles that
are inherent ("mother" from the "mother" frame).  Another problem
is that I believe my frames are too specific.  I have different roots
for "teaching" and for "studying", so that the word for "teacher"
and "student" come from separate frames, and separate triconsonantal
roots.

So, then, if you wanted to create Ideal Megdevi, what you would
probably want to do is expand the number of agent roles, and
make them much more specific.  You might also define different
types of triconsonantal roots (e.g., is the realm of "teaching/studying"
different in some fundamental way from the realm of "parentage/
birth"?).  The different types of roots, then, might use the different
types of agent roles in different ways--but, crucially, in each case
the use would be specific.  (This, for example, would allow for
my "natural nouns".  You could use the same agent role pattern,
but if the triconsonantal roots were somehow marked as being
different, then the user would know that using the "natural noun"
in the X-type frame would mean an agent, and using it in the Y-type
frame would mean an inherent role held by a human not necessarily
associated with an action).  As for verbs, again, the way a verb
works can be different depending on the type of triconsonantal
root.

Before going any further, if you don't want to mark triconsonantal
roots differently, there can simply be a ton of agent roles, and not
all will be used for every triconsonantal root.  This is certainly true
of Arabic and Megdevi: not all possible combinations are licit words.
In this case, you'd have to make your verbs very specific.  If in the
birth frame there is no meaning you can associate with "X performs
upon Y for the benefit of Z", then it has no meaning.  If it does have
a meaning, then it does.

 From the point of view of the language creator, I really felt
different when creating words for Megdevi.  Rather thinking,
"Okay, I want a word for 'doctor'," I thought, "Here is my
root for 'healing'.  How would I translate this word into English?"
In other words, I was creating the words, and they had a kind
of prototypical meaning, and I would try to think of how that
concept could be translated into English.  Sometimes it translated
into a single word; sometimes it didn't.  The problem you might
have going the other way is that users will be looking for "words".
I don't want to say anything Sapir-Whorfian, but if I need to
translate something, I look for the word I need to translate the
word I want translated; I don't think in concepts.  It will take
practice to switch your thing, and also to use the system to its
fullest extent.  For example, in Arabic you can't really make up
words using roots and patterns; in a conlang, why not?  You
should be able to, and it would seem like that would be the best
way to use that type of a language.  To expect a user to be able
to pick that up without practice right away, though, is expecting
a bit too much.

Anyway, an interesting project.  Let us all know if you do anything
with it.

-David
*******************************************************************
"A male love inevivi i'ala'i oku i ue pokulu'ume o heki a."
"No eternal reward will forgive us now for wasting the dawn."

-Jim Morrison

http://dedalvs.free.fr/