Semantic Requirements Engineering - Intentional Perspectives on Information Systems Engineering - page 72

method engineer is suggested to add a relationship on Message to keep semantic
consistency of the meta model.
By establishing a semantic mapping frommeta model elements into the ontology,
we can avoid producing semantically inconsistent meta models [ 2] . The meta model
ontology has several inference rules to keep semantic consistency on meta models.
In this example, the used inference rule is “If concept A and B are mapped into MA
and MB in the meta model ontology and there is a relationship MRAB between MA
and MB, then the relationship that can be mapped into MRAB exists in the meta
model.”
Note that we should explain the stereotypes attached to the domain ontology
in Fig. 3. These stereotypes result from the names of concepts of a meta model
ontology to clarify the relation between the domain ontology and the meta model
ontology. For example, an ontological concept Stop has a stereotype Event in Fig. 3
and the type Event is the same as the concept Event of the meta model ontology
shown in Fig. 4.
3 Application to GORA
Goal-oriented requirements analysis (GORA) methods are one of the promising
approaches to elicit requirements [19, 23] and are being amplified so as to put them
into practice [ 14] . In this approach, customers' and users' needs are modeled as
goals to be achieved by a software-intensive system that will be developed, and the
goals are decomposed and refined into a set of more concrete sub-goals. After fin-
ishing a requirements elicitation process, the analyst obtains an acyclic (cycle-free)
directed graph called goal graph . Its nodes express goals to be achieved by the sys-
tem, and its edges represent logical dependency relationships between the connected
goals. We have two types of goal decomposition; one is AND decomposition and
another is OR. In AND decomposition, if all of the sub-goals are achieved, their
parent goal can be achieved or satisfied. On the other hand, in OR decomposition,
the achievement of at least one sub-goal leads to the achievement of its parent goal.
Root goals, having no parents in a graph, expresses the needs that the customers
would like to fulfill ultimately and the analyst tries to achieve them by combining
sub-goals. Figure 5 shows an example of a goal graph to elicit requirements of a seat
reservation system of trains, which is a part of a screenshot generated by our GORA
supporting tool [ 16] . In this figure, a root goal N1 “Seat reservation system required”
is decomposed into three sub-goals N2, N30 and N4 in AND-decomposition. The
arc crossing over edges shows AND decomposition.
In order to construct a goal graph of semantically higher quality, knowledge of
a problem domain where the system to be developed operates, so called domain
knowledge, is necessary. In Fig. 5, stakeholders and/or an analyst, who are con-
structing the goal graph, need the knowledge of a reservation business domain and
a train service domain, etc.
The example of using our ontological technique is the application of an ontol-
ogy as a source of domain knowledge for requirements elicitation in the GORA