Like object technology, knowledge technology originates from a different
view on software. Most software nowadays consists of programmes executing
prescribed procedures or algorithms. Programmes can therefore only support
and execute well-structured processes. These procedures must be exactly
prescribed in advance. The order in which the user performs his actions
is also prescribed. The behaviour, in other words, of a well-structured
COBOL programme. Object technology in combination with a graphical user
interface does away with the necessity of formally prescribing the order
and coherence of the actions. Each function an object performs, however,
remains a form of formally prescribed behaviour.
Knowledge technology focuses on the development of
applications that simulate or support matters such as human reasoning,
patter recognition and the human learning process. These are all applications
with weak structures. Neither the order of the actions nor the way of
acting are known exactly in advance.
Reasoning Systems
An important area of application for knowledge technology is the development
of applications that simulate or support human reasoning. This kind
of applications was originally facilitated by declarative programming
languages such as LISP or PROLOG. With these languages it is possible
to construct systems with reasoning rules, for instance with regard
to diagnosing illnesses. The user presents a problem to the system,
for example the symptoms a patient shows. On the basis of the problem,
the system applies the right reasoning rules and arrives at a solution,
consisting in this case of the probable diagnosis.
Nowadays, so-called hybrid development environments are used in knowledge
technology. These are environments in which the developer can develop
in a procedural, a rule-based and an object-oriented manner all at the
same time.
An example of an integrated application is a module for the acceptance
of insurance policies, developed for a large number of insurance companies.
The insurance agent, who wants to hand in a policy application with
a certain company, works with his own, usual software. Behind the scenes,
a knowledge-based technology module assesses the application and, if
acceptance is in order, processes it immediately at the agent's workstation.
The policy, the premium calculation and the invoice are made immediately
after acceptance of the policy. Knowledge technology has reduced the
process of acceptance from several weeks to several minutes.
Neural Networks
Within the field of knowledge technology, an increasing amount of different
technologies are used besides reasoning programmes, for example neural
networks and genetic algorithms.
These are forms of so-called fuzzy logic.
Neural networks imitate the working of the human brain. The brain consists
of a network of interconnected brain cells (neurons) that are together
capable of learning, remembering and especially recognising things.
Artificial neural networks are mainly applied for solving problems to
do with pattern recognition, such as the recognition of handwriting
or voice recognition.
The knowledge in a neural network is not programmed but trained. A
neural network only recognised a signature after 'seeing' it for a certain
number of times and after learning from the user which person belongs
with it. This training can, by the way, be done by the computer. To
compensate the disadvantage of the time this training takes, certainly
at first, there is the advantage that it is easy to add new knowledge
to the network (a new signature, for example) by means of follow-up
training.
An example of a neural network in combination with a knowledge technology
system is a system supporting unemployed people in finding a job. On
the basis of the profile of the unemployed person and the market situation,
the neural network estimates the expected duration of the unemployment.
Genetic Algorithms
Genetic algorithms are self-learning algorithms. The idea is, for example,
to have algorithms simulate a certain real-world situation as well as
possible. In doing so, they are not only capable of representing the
past course of events, but also of calculating the expected future course
of events. One starts out with a certain set of algorithms, and after
a more or less artificial selection of algorithms the set is found which
best represents the behaviour of the problem situation. The finding
of the right algorithms strongly resembles a biological evolution process
with phenomena such as selection and cross-over and resulting in the
'survival of the fittest' of algorithms.
An example of an application that uses genetic algorithms is an application
in which genetic algorithms are applied to the results of a credit score
system in order to determine which algorithm best predicts a client's
chance of failure.
Knowledge Management
To a certain extent, developing knowledge-based systems like reasoning
systems, neural networks and genetic algorithms is already a form of
knowledge management. After all, knowledge-based systems offer the possibility
to record, distribute and maintain certain knowledge and experience
of individuals within an organisation.
The grip an organisation thus gets on knowledge grows, and people can
(strategically) determine where, when and how the knowledge shall be
used.
When referring to knowledge management systems, however, one rather
means systems that support the collective knowledge and experience of
an organisation. Employees from various departments and disciplines
can access a central knowledge base, which supports the execution of
various tasks. This knowledge base is a database containing
data and objects that support the users in exchanging knowledge.
Important with this type of systems are the management and the maintenance
of the data relating to the knowledge of the users. Large organisations
often appoint a knowledge manager to handle these tasks.
A knowledge management system was developed for one of the Dutch government
departments. The application is used as a central working memory for
lawyers. Through the system they have access to data that may support
them in the execution of new tasks. The application uses knowledge technology
in combination with retrieval techniques such as hypertext and full
text retrieval. Knowledge technology was applied here for the classification,
interpretation and conceptualisation of data within the knowledge domain.
Besides this, 'knowledge technological' interview techniques were used
to extract the knowledge.
Conclusion
In the coming years, knowledge technology will become an essential part
of information
technology, primarily as a technology to model and automate less
structured tasks such as assessment and diagnosing, and secondly to
manage stored data that support people's
knowledge and to keep these data accessible. Such knowledge management
systems are indispensable to keep the enormous amounts of distributed
digital data accessible in the future. In the long run, the
information superhighway will lead to the existence of a world-wide
network of computers on which we will store and manage much of our knowledge
in the form of digital data. Knowledge
technology will be essential to be able to realise this in an effective,
beneficial way.