KAW: Knowledge Acquisition,
Modeling and Management
Sisyphus Projects
Background
One major dimension of research reported through KAW is that
theories, methodologies and techniques have been made operational
through their implementation in tools that have been demonstrated to
the community.
In the early years it was natural also to attempt to share these
tools in order to evaluate their efficacy when used by those other
than their originators. In the early years the attempts to share
tools were not very successful because:-
- Porting tools to different environments proved difficult
- Integrating tools from other sites proved difficult
- Supporting multiple sites was problematic
- Method of use was not well-defined by originators
- Tools were effective for orginators but ineffective for others
These issues led to the development of the Sisyphus series of
challenge problems in which, rather than share tools, a knowledge
acquisition problem was defined and tool developers were challenged
to solve it with their tools.
Sisyphus Problems
Four Sisyphus problems have been defined:
- The first Sisyphus project was a room allocation problem, a
resource allocation task in which a number of people with
differing requirements as to type of room have to be allocated
appropriate rooms from a number of rooms with differing
characteristics. It was based on an ESPRIT study at the GMD in
Germany and intended as a knowledge acquisition challenge problem
that did not seem amenable to many of the existing knowledge
acquisition techniques that focused on heuristic classification. A
set of papers on Sisyphus-I solutions was published in a special
issue of the International Journal of Human-Computer Studies 40(2)
1994. An article on an example solution is
available on the web.
-
- The room allocation problem proved to relatively simple and
the second Sisyphus project attempted to provide a realistic
knowledge-engineering problem. It was based on the domain
knowledge for elevator configuration that was used to build
Marcus' VT system. A set of papers on Sisyphus-II solutions was
published in a special issue of the International Journal of
Human-Computer Studies. An article on an example solution is
available
on the web. The Knowledge Systems Goup of the Knowledge Media Institute at the Open University in the UK has
also made an operational solution available interactively through the web.
-
- The principal objectives of the Sisyphus III project are:
- To provide for better quantitative comparison of KB systems
and the methodologies employed to build them, through use of a
set of achievement metrics
- To provide more realistic access to actual KA material in a
staged series of releases
- To obtain more complete records (or knowledge engineering
meta-protocols) concerning the processes that the knowledge
engineer goes through in the KBS construction process
Some Sisyphus-III studies will be reported at KAW98.
-
- In recent years many knowledge acquisition tools have been ported to operate through
the Internet and World Wide Web and it has become possible to provide access to tools to
the community in a way that alleviates the problems noted in the first section above.
Sisyphus-IV is a project to encourage projects collaborative use of tools through the net
and web and by the integration of web tools at different sites through the net.
Some Sisyphus-IV studies will be reported at KAW98.
gaines@cpsc.ucalgary.ca, 18-Nov-97