parsing documents vs. human assistance (was Re: [om-list] hello again
...)
Luke Call
lacall at onemodel.org
Sun Sep 1 06:17:21 EDT 2002
Rajeev,
It seems that you and Tom have both expressed both humility and wisdom.
I really like associating with such persons and hope to emulate that.
I'll think you'll find that Mark and Lee have similar traits.
I completely agree with you that parsing human grammar to populate a
knowledge base is not likely to be a successful path very soon, and I
imagine that to be true for even longer than you suggest. So I think
that the first path would be to make it easy for humans to manipulate
the model, interactively and collaboratively, then add automation tools
as ability and opportunity present themselves--with an eventual goal of
populating it based on all available human text. And I also agree that
it may likely need an assistance from humans to be able to process text
as input.
I glanced at the URLs you listed and anything along those lines would be
helpful (xml too), but I think maybe we first need to put the basics in
place. I have a rudimentary system building now, and unit tests, but
need to add some more features so it does something (barely) useful,
before it's worth sharing.
My first iteration would be just to create a simple model allowing one
to create ad-hoc entities and to relate them to each other and assign
properties. Then to be able to share that with others' work (both coding
on the project and entering data in the system), and see where it goes
from there. Very basic at the beginning, and this first pass should make
it much easier to communicate with others about what I have in mind and
how to build it together.
I'm certainly open to work on the web site as well; what have you got in
mind? Any help is appreciated.
You have a lot of good ideas; I will try hard to get something out that
is a beginning (of what I envision, anyway), and then we can build on
that and add things along various lines that folks here discuss (or
discard it and do better), and we can work together. This is all good.
Thanks!!
Luke
More information about the om-list
mailing list