Information Technology Reference
In-Depth Information
4.
Nurture
: using the interfaces that they drew, the end-users are invited to
provide data examples that are analysed to infer and arbitrate possible con-
straints and dependencies;
5.
Bind
: the pre-integrated schema is completed and refined into a non redun-
dant integrated conceptual schema;
6.
Objectify
: from the integrated conceptual schema, the artefacts of a proto-
typical data manager application are generated;
7.
Wander
: finally, the end-users are invited to play with the prototype in order
to refine and ultimately validate the integrated conceptual schema.
In order to position end-users as major stakeholders throughout the data require-
ments process, the approach uses form-based interfaces as a controlled basis for
joint development, analysis and discussion. In particular, in order to make the
development of the interfaces more accessible and focus the drawing on the sub-
stance rather than (ironically) the form, the available graphical elements are
restricted to the most commonly used ones (
forms
,
fieldsets
and
tables
,
in-
puts
,
selections
and
buttons
) and limited the layout of forms as a vertical
sequence of elements, which also simplifies the transition from the form model
and the ER model. This drawing phase is supported by the RAINBOW Toolkit,
which is the dedicated and integrated tool support intended to assist end-users
and analysts during the different RAINBOW processes.
The interfaces being drawn by non experts and possibly multiple end-users
increases the possible inconsistencies among the individual labels and the struc-
tures used in the forms [2]. The semantic and structural similarities are therefore
analysed to manage commonality and standardise the form constructs and their
underlying data counterpart. Semantic similarities arise due to the richness of
written natural language, which can lead to spelling and meaning ambiguities.
Structural similarities which occurs when two entity types share a pattern, which
is a bijection between two sets of attributes belonging to different entity types.
RAINBOW deals with the elicitation and subsequent unification of semantic
similarities using String Metrics, Ontologies and dictionaries. Though the forms
and their underlying data models have a tree-like arborescence, their structure
is simple and does not necessitate using complex techniques such as tree min-
ing approaches to discover structural similarities. The shared patterns are in-
stead elicited by comparing the different entity types, and the structures are
subsequently unified according to the meaning of the pattern (equality, union,
comprehension, complementarity, composition or difference).
The RAINBOW approach also deals with the integration of each schema cor-
responding to a form into a single normalised schema representing the domain
of application, before generating and testing the resulting associated applica-
tive components. However, before leading these processes, an important step
consists in eliciting the relevant constraints and dependencies of the domain of
application.
