Modeling Design Patterns with Description Logics: A Case Study - Advanced Information Systems Engineering - page 176

Moreover, providing mapping between the pattern and the problem is a critical fea-
ture to support designers in applying the patterns to resolve their problem.
Here is a fragment of the representation of the problem-at-hand in Fig. 1 in terms of
concept and role instances in the ABox:

- Role(Team Sector)
- Role(Executive Controller)
- Role(Planning Controller)
- Agent(Bob)
- Goal(Ensure traffic safety in its sector)
- Goal(Manage traffic in sector)
- Goal(Manage inbound traffic)
- Goal(Resolve traffic conflict)
- play(Bob, Executive Controller)
- play(Bob, Planning Controller)
- isPartOf(Executive Controller, Team Sector)
- isAndDecompositionOf(Ensure traffic safety in its sector, Manage traffic in sector)
- isAndDecompositionOf(Ensure traffic safety in its sector, Manage inbound traffic)
- hasPosContribution(Manage inbound traffic, Manage traffic in sector)
- provide(Executive Controller, Resolve traffic conflict)
- DelegationOnExecution(Del-exec1)
- hasDelegator(Del-exec1,Team Sector)
- hasDelegater(Del-exec1,Executive Controller)
- hasDelegatum(Del-exec1,Manage traffic in sector)

4.5 System Architecture

Fig. 6 depicts the architecture of our implemented system. Though this work supports
two types of queries (SPARQL and SQWRL), most system components and artifacts
are common for both inputs (normal line). The ones with thick lines refer to parts for
SPARQL, while dashed lines to SQWRL. In both cases, the implemented system re-
quires the same input SI* model representing the problem-at-hand and a set of SI* mod-
els representing patterns.
Since we need some inference capabilities to deduce implicit facts, we use a rule
engine (i.e., JESS) that is integrated in Protege. Essentially, a rule engine takes an input
(rules and facts) and produces a model. In SQWRL setting, the input consists of the
TBoxandABoxdefinedsofar,patternsinSQWRL,andSWRLrules 3 . In this sys-
tem, the model (produced by the rule engine) contains the resultset of the matching. In
SPARQL setting, the input to the rule engine only contains TBox, ABox, and SWRL
rules. The rule engine produces a model containing inferred knowledge from available
facts and rules. Using the Model-to-OWL library in Protege, inferred knowledge is
added to the knowledge base (TBox and ABox). By means of the OWL-

DL

reasoner
(e.g., Pellet, JENA), we can query the revised knowledge base to find a match to a pat-
tern. In both settings, if the length of resultset is zero, then there is no match found in

3

available at: http://www.w3.org/Submission/SWRL/swrl.owl