Information Technology Reference
In-Depth Information
requirement perspective. The rationale behind
this approach is as follows.
Firstly, in terms of modeling philosophy, in
ontology modeling a concept is described by add-
ing its details and implicitly defining in a bottom-
up fashion the scope of the concept through the
details. Whereas, in feature modeling, a concept
is described by first setting its scope and hierar-
chically adding its details in a top-down fashion
(Czarnecki et al. 2006). This feature is quite
interesting as it allows the context modeler to
devise, in a top-down fashion, generic and reus-
able context features which can be shared among
all applications that need to use this context. The
relationships between context features express
the context variability from the application point
of view.
Secondly, according to the context working
definition previously presented in “Context-Aware
Systems,” we consider that the context knowledge
is composed of a set of small contextual knowledge
pieces namely context primitives which include
context entities, attributes, associations, and rules.
Each context feature corresponds to a specific
set of context primitives. The focus is a concept
representing the point of view the application is
interested in looking at the current context. Each
focus corresponds to a specific set of context
features. Given a focus, a relevant subset of these
pieces will be used to generate a per-application
customized contextual knowledge. Obviously,
considering only the relevant context primitives
will improve the reasoning performance and
reduce response time which is a vital issue in a
pervasive environment.
Thirdly, as developers usually do not have full
understanding of the context internal semantic,
“promoting” the context information using the
feature model will enable the contextual knowl-
edge visibility from different views in a top-down
fashion. Another advantage is that these context
features might be shared between applications
which significantly enhances the reusability of
context information and reduces application
complexity.
The Conceptual Meta-Model
for Context Management
We import the concepts of features from FODA
(Feature Oriented Domain Analysis) (Kang et al.
1990). FODA appeals to us because features are
essential abstractions that both context consumer
and provider understand. Thus, the main concept
in the feature description language FODA is the
feature itself. Here a feature is a set of context
primitives that is relevant to some stakeholder
from a specific “focus” point of view. Figure 2
depicts the proposed conceptual metamodel. The
concepts of the conceptual metamodel were identi-
fied and grouped into two different sections: the
context related concepts (white), and the context
features concepts (shaded).
The main construct for representing context
knowledge is the
ContextPrimitive
which repre-
sents the base context constructs (primitives)
mentioned above: entity classes, entity attributes,
entity associations, and rules. Two types of rules
could be identified: (i) Consistency rules provide
a mechanism for context consistency by specify-
ing conditions that must be held in the context
information. For example, a consistency rule could
specify that if the person is cooking, she must be
in the kitchen. (ii) Inference rules used to gener-
ate new context information after reasoning on
the existing one. For example, an inference rule
could conclude that a person is sleeping if the
light is off and the time is night. Further modeling
constructs are axioms that add additional facts
about the entities and attributes. These are: spe-
cialization and equivalence relationships that may
be specified between two entity classes, two at-
tribute classes, or two association classes.
Search WWH ::
Custom Search