|
Main
Description
Status
Research
Publications
Demo
People
Funding
Links
|
Status
Our initial evaluation of KANAL as a stand-alone system shows its usefulness
in finding errors and suggesting fixes. It also has been
successfully integrated within an end-to-end system that support users
entering domain knowledge as a part of DARPA Rapid Knowledge Formation (RKF)
program.
KANAL is a tool for building and checking process models. By
relating different pieces of information in process models among themselves
and to the existing KB, it performs a variety of verification and validation
checks and propose useful fixes.
We performed an initial evaluation in December 2000, which has been reported in
Kim and
Gil, IJCAI-01. The results show that KANAL can find most of the errors in
the process models and suggest useful fixes including the fixes that directly
point to the sources of the errors.
KANAL has been successfully integrated within an end-to-end system (called
SHAKEN) that supports subject matter experts entering domain knowledge as part
of the DARPA Rapid Knowledge
Formation (RKF) program. During summer 2001 and January 2002, the SHAKEN
system has been evaluated by lightly trained biologists where they author
biology textbook knowledge using the system. The resulting knowledge bases
answer questions that a student might be asked based on the
chapter's contents.
We have recently built an Action Editor in order to support interleaving of process
model authoring and checking while extending background knowledge of actions.
Also we have refined the KANAL interface based on feedback from end users.
The improved system was evaluated by military officers in describing
and checking their plans (courses of actions) during Fall 2002. The details
are desribed in Kim and
Blythe, IUI-2003.
|
