|
An Expert System (ES) is
a computer program that performs difficult and specialised
tasks at the level of human expert. The structure of the ES
presents two main independent modules: the Knowledge Base
(KB) and the Inference Engine (IE). The KB contains the
overall knowledge of the process (i.e., in our case, stream
nutrient retention) usually codified by means of heuristic
rules (a rule is a set of conditions and conclusions linked
for a given hypothesis). The IE is the software that
controls the reasoning operation of the ES chaining
optimally these rules. The ES has also additional
connections to process data and external applications,
together with its ability to perform some kind of
explanation or justification through computer interface to
inform the user. The bottleneck of the ES development is the
knowledge acquisition, both general (related to the domain)
and specific (distinguishing features related to the site
where the ES will be applied). While the general knowledge
of the process can be extracted from literature and human
experts, the specific knowledge can only be acquired by
processing empirical data directly collected from each study
site.
The
overall knowledge acquired must be structured and codified
by means of a set of "if-then" rules (IF
<conditions> THEN <conclusions>) as shown in
this example:
IF
<Nutrient comes mainly from non-point sources>
&
<Denitrification is the main mechanism for nutrient
removal>
THEN
<Create those conditions in the stream that enhance
denitrification>
To
build up the Knowledge Base we propose the following
steps:
- To review,
acquire, and structure the existing knowledge from
the literature and from direct experiences in stream
management. This review offers a wide range of potential
causes for observed trouble shooting, along with a
correspondingly wide range of potential solutions.
Notwithstanding, this general knowledge is really scarce
and must be exhaustively complemented.
- To arrange
a set of interviews and workshops with the stream
managers (in our case, the end-users) and the
scientists who are carrying out the experimental testing
in the different scenarios. Though this heuristic
information is very valuable to define the fundamentals
of the ES, experts tend to emphasise recent, first-hand
experience, providing a limited and potentially biased
view of the stream water quality problems.
- To acquire
all the knowledge from each specific scenario. This
work will be based on the available database (this is
one of the supporting reasons to choose the scenarios
that are contemplated in the project) and on the results
obtained from WP1, WP2, and WP3. In this sense, we are
familiar with some machine learning methods (e.g.
Linneo+ and ID3), which will let us examine the
relationships between stream nutrient retention and
several physical, chemical and biological parameters.
- Once there
is a global agreement among the different partners
involved in the project about the optimal management
strategies to be proposed by the ES under each
particular scenario, all this knowledge will be codified
by means of "if-then" rules to conform the KB
of the ES. This is a long and iterative process,
mainly supervised by the End-Users during several
meetings. During this process, some partial advances
will be tested using a commercial shell from Gensym
Co.(G2).
The development
of the ES prototype will be carried out on a platform
that will be absolutely codified by our own programmers.
This is the best option to obtain user-friendly software,
directly adapted to the needs of end-user, and able to
integrate empirical knowledge with knowledge about local
environmental conditions and its possible relationship with
the global environment, with judgmental knowledge about
human beliefs, intentions, desires and priorities, and with
theoretical knowledge about biological, physical, and
chemical phenomena.
We
also propose that a most effective management decision
system would result from integrating the ES with spatial
information within the context of a GIS system (e.g.
ArcView). This is our expected last step in the development
of the prototype proposed throughout the project.
Finally, a global validation of the prototype in
different scenarios chosen by the End-Users will be done. |
Web curator |
