Discovering Critical Factors Affecting RDF Stores Success
203
data models following the approach provided by the INTER-IoT project [12, 22],
which defines mapping rules to perform semantic translation from these models to
another model which plays the role of central hub to enable shared understanding.
3.3.5 Zero Impedance Mismatch
Even the SWT are enabling technologies for information integration, they are affected
by the issue of the Impedance Mismatch, i.e. the discrepancy both in language syntax and semantics between the models of the programming languages and the RDF
semantic models [6]. In traditional relational-database applications, various solutions
have been implemented in order to bypass the Impedance Mismatch. In particular,
the object-relational mapping (ORM) software solutions (such as Hibernate) offer
programmatic access towards relational data sources through the mapping of the
data objects into programmatic objects, thus allowing developers to programmatically handle data exploiting their habitual application programming interface [27].
However, this approaches is not applicable to Semantic Web based applications since
the conceptual model represented through the SWT languages differ substantially
from the relational model and from the object-oriented model which characterizes the
ORM. For this reason, it is essential to identify new strategies in order to provide an
immediate programmatic access to Semantic Web data, thus limiting or eliminating
the Impedance Mismatch [27].
3.3.6 (Distributed) Reasoning
It is the capability of a specific component (the reasoner) to derive new knowledge
directly from assertions contained in the store, applying an appropriate set of logical
rules [16]. The deductions that follow can be used with the dual purpose of extending
the basic knowledge available for queries and calculations, enlarging the network of
concepts, and validating the knowledge base itself. Often the inference capability
is integrated directly into the RDF store and is transparent to the users. In other
applications, inference is implemented as an external tool that is manually initiated
and the entailments that follow are manually added to the store. The latter approach
can be used as a way of mitigating the computational cost of performing inference
if it negatively impacts the overall performance of the store. In fact, the scalability
is one of the main barrier of currently existing reasoning systems. Thus, one of the
aim of researchers is to overcome this obstacle, exploiting, for instance, the benefits
of parallel computation techniques applied to reasoning algorithms. In addition, an
approach based on distributing reasoning can improve performance reasoning with
large data sets. Many of the existing RDF stores perform inference using rules-based
reasoning engines. When evaluating reasoning engines, it should be checked keyfeatures such as the compliance with standard languages for reasoning (e.g. OWL
2 RL, OWL 2 QL, and OWL 2 EL), the inferencing strategies (forward chaining or
backward chaining) and the life cycle of the inferred tripled (they are materialized or
203
data models following the approach provided by the INTER-IoT project [12, 22],
which defines mapping rules to perform semantic translation from these models to
another model which plays the role of central hub to enable shared understanding.
3.3.5 Zero Impedance Mismatch
Even the SWT are enabling technologies for information integration, they are affected
by the issue of the Impedance Mismatch, i.e. the discrepancy both in language syntax and semantics between the models of the programming languages and the RDF
semantic models [6]. In traditional relational-database applications, various solutions
have been implemented in order to bypass the Impedance Mismatch. In particular,
the object-relational mapping (ORM) software solutions (such as Hibernate) offer
programmatic access towards relational data sources through the mapping of the
data objects into programmatic objects, thus allowing developers to programmatically handle data exploiting their habitual application programming interface [27].
However, this approaches is not applicable to Semantic Web based applications since
the conceptual model represented through the SWT languages differ substantially
from the relational model and from the object-oriented model which characterizes the
ORM. For this reason, it is essential to identify new strategies in order to provide an
immediate programmatic access to Semantic Web data, thus limiting or eliminating
the Impedance Mismatch [27].
3.3.6 (Distributed) Reasoning
It is the capability of a specific component (the reasoner) to derive new knowledge
directly from assertions contained in the store, applying an appropriate set of logical
rules [16]. The deductions that follow can be used with the dual purpose of extending
the basic knowledge available for queries and calculations, enlarging the network of
concepts, and validating the knowledge base itself. Often the inference capability
is integrated directly into the RDF store and is transparent to the users. In other
applications, inference is implemented as an external tool that is manually initiated
and the entailments that follow are manually added to the store. The latter approach
can be used as a way of mitigating the computational cost of performing inference
if it negatively impacts the overall performance of the store. In fact, the scalability
is one of the main barrier of currently existing reasoning systems. Thus, one of the
aim of researchers is to overcome this obstacle, exploiting, for instance, the benefits
of parallel computation techniques applied to reasoning algorithms. In addition, an
approach based on distributing reasoning can improve performance reasoning with
large data sets. Many of the existing RDF stores perform inference using rules-based
reasoning engines. When evaluating reasoning engines, it should be checked keyfeatures such as the compliance with standard languages for reasoning (e.g. OWL
2 RL, OWL 2 QL, and OWL 2 EL), the inferencing strategies (forward chaining or
backward chaining) and the life cycle of the inferred tripled (they are materialized or