On Roles of Models in Information Systems - Intentional Perspectives on Information Systems Engineering - page 26

we would like to have is a modeling language which supports increased formality
of expression through systematic addition of specification detail, starting from an
informal basis. An example would be that an initial informal natural language speci-
fication is systematically enriched with more and more detail until the most detailed
specification can be transformed into a formal form, based on linguistic analysis.
Another example would be a graphical language for, e.g., information modeling,
which is as suitable for high level informal sketches of information objects as it is
for describing detailed data structures.
An approach to formalization of natural language specifications is found in an
extension to fuzzy set theory which is called computation with words [38] . The
basic idea is to quantify words and sentences so that they can be given mathemati-
cal treatment. Another approach is to relate model fragments (semantic frames) to
natural language sentences, see, e.g., [ 39] for a collection of semantic frames.
Graphical languages are usually used for sketching systems structures on the
exploratory road from informality to formality. These languages are widely used,
e.g., UML, ER-diagrams. Graphical languages for specifying every detail in a struc-
ture are far between [ 10] . A comprehensive analysis of graphical languages has
recently been published in [17] .
3.5 Comprehension
With the penetration of computers into all realms of human affairs comes the need
for understanding what the computers do, to enable people to judge whether the
software produce correct data. Both data specifications and software specifications
must be comprehensible in the sense that they can be understood and validated by
people that are not computer specialists.
The need for comprehension of system specifications is complicated by the
necessity that somewhere in the development process system specifications are
stated in executable terms, that is, in the languages of data bases and computer
software, which are usually only understood by specialists.
A solution of the dilemma has been sought in model driven development (MDD),
also known as model driven architecture [ 43] and model driven engineering [ 44] .
The main idea of using the term model driven is that, in order to increase comprehen-
sibility, software and data specifications are developed in a language that reflects the
particularities of the domain. These specifications should later on be automatically
translated into executable software.
The difficulty lies in the details. The current state of the art of expressing exe-
cutable specifications requires the specification of so many details of the data system
realm that the comprehensibility of the specifications suffers severely for those that
are not software specialists.
3.6 Validation
Systems are composed of subsystems which are related to each other. The
relations among the subsystems are called the system's structure. A concrete