|
by Armin
Eberlein,Dale Fukami &
Wonh Chieh Lam
Start ý Method
ý Implementation ý Results
ý Sources and PDF
RESULTS
The current results are still preliminary
because the interface layer is being refined. However, initial results
are promising. The methodology is completely implemented in the Telos
language and, together with the intelligence models, provides comprehensive
guidance to the developer. The expert system can be asked what further
work is needed for a certain requirement or a requirements document
and responds to the user with instructions for the tasks that are
necessary to further develop the system specification according to
the implemented methodology. The user also is informed about any actions
that violate constraints specified in the database, preventing incorrect
use of the development methodology. This shows that the initial objective
of the research is met and intelligent support for the traditionally
difficult first phase of the system development life cycle can be
provided using an expert system.
Requirements management issues are addressed
by the links that are created among the requirements. This allows
the tracing of a requirement back to its initial source and caters
to the impact analysis of a change request.
EVALUATION
In order to evaluate the approach, a
case study is planned. Connections to the telecommunications industry
are already established, and the experimental development of an intelligent
network service is intended. The framework for requirements engineering
has already been adapted to the intelligent network (IN) architecture
by mapping the system level of the framework to the service level
of the architecture, the high-level components to the service features,
and the low-level components to the service-independent building blocks.
This customization offers more possibilities for intelligent support.
The case study is expected to reveal
any shortcomings of the overall approach of using an expert system
for the early system development life cycle and offers the possibility
to mature the methodology. Scalability is also of concern and modifications
might become necessary to allow the development of large-scale systems
using this expert system. Currently, this is one of the major issues.
The use of the ConceptBase tool for requirements engineering can easily
result in a number of objects that drastically slow down database
queries. Although the current experimentation still performs well,
a large-scale case study will require a high-performance workstation
to ensure acceptable response times and usability.
WHAT WEýVE LEARNED
The requirements engineering methodology
and its implementation in a tool aptly portray the use of an expert
system to facilitate the gathering of requirements information. The
tool provides active and passive guidance at various levels to the
system developer, resulting in a high-quality specification of the
new system. This will result in faster time-to-market and increased
reliability and customer satisfaction.
I encourage the expert system community
to use expert systems not only for specific domains, but to apply
them to software development as well. The software community is still
struggling with many difficulties and would welcome any support for
developing reliable software that meets customersý needs. Widening
the scope, the general use of artificial intelligence during system
development is an area of research that requires much more work, but
has great potential for success.
PREVIOUS
NEXT
Circuit Cellar provides up-to-date information for engineers. Visit
www.circuitcellar.com for
more information and additional articles.
For subscription information, call (860) 875-2199, subscribe@circuitcellar.com
or subscribe online.
ýCircuit Cellar, the Magazine for Computer Applications. Posted with
permission. |