Database Reference
In-Depth Information
Figure 5: Information represented by means of a class diagram and a data fl ow diagram
for the sentence, “Patient is assigned to Room. Room belongs to Ward.”
to a room (the phrase “is assigned to”) has been lost, and this has been replaced by an as-
sociation class, rewriting the problem in static code. In Figure 5(b), on the other hand, the
relationship between patients and rooms is lost, and the information gathered is transcribed
functionally, which, ultimately, indicates merely that the patients have some kind of attribute
that is updated on the basis of wards and rooms.
Moreover, the intrinsic ambiguousness of the concept map enables the categorization
to be carried at the end of the analysis, when all the relevant information has been acquired
and correctly conceptualized. Accordingly, in the concept map shown in Figure 3, there
is no constraint that demands that “patient” be modeled in advance as a class or external
entity or “is assigned to” as a relationship or a process. This decision is postponed until the
problem domain is well enough understood and the best-suited development paradigm has
been selected. This view of analysis is vaguely similar to the one proposed by Ceri (1983)
and Mayr and Kop (1998), who also use generic representation formalisms to record the
problem domain information before going on to create the conceptual models.
By modeling the problem according to the concept map representation schemes, we
have mapped set P to set PM. It remains, therefore, to defi ne the correspondence between
the sets PM and CM.
Determining Conceptual Model Fitness
As mentioned above, the correspondence between PM and CM is established by compar-
ing the generic conceptual model of each problem P (or the respective concept map) with the
conceptual models used by the methods and techniques. Being based on distinct theoretical
foundations, however, the concept map and the classical conceptual models, such as the
class diagram or the data fl ow diagram, cannot be compared directly. On the one hand, the
concept map records ambiguous information, which is, therefore, susceptible to different
interpretations. On the other, the conceptual models, albeit to different extents, record the
information on the problem using a strict semantics and, therefore, a single meaning.
Owing to this impedance mismatch, the theoretical foundations of the concept map
and the conceptual models need to be approximated, that is, assimilated. This is achieved by
disambiguating the concept map. The ambiguity of the concept map is removed by ascrib-
ing a given interpretation to each concept and association in the concept map; that is, each
