Table 13 compares several aspects of data integration frameworks in open environment. The proposed model is compared to two various frameworks in the literature. We will be comparing the proposed model to the two frameworks proposed in (Y. D. Wang 2009) and (Xue 2010).
As shown in Table 13. The proposed model is intensional in nature. This is due to the class of logic based on which the system is modeled. Both the systems presented in (Y. D. Wang 2009) and (Xue 2010) are based on an extensional reduction model. It has been illustrated in Chapter 3 that the extensional reduction model does not adequately describe a conceptualization. It has also been demonstrated that information systems, in general, and open environments, in particular, are intensional in nature. As such, the use of intensional logic and an intensional model are natural choices for data integration in open environment.
Another important aspect shown in Table 13 is the dynamic nature of open environment. Neither (Y. D. Wang 2009) nor (Xue 2010) addressed the dynamic nature of open environment. The proposed framework, however, addresses the dynamic nature of open environment through the use of an intensional logic. The proposed framework also models a mediated P2P network, which has N number of peers, with 2N IEL theories. This enables the system to continue to function while peers adapt to the changes in their local networks.
Table 13 also highlights that the proposed model uses ontologies as the source of semantics as opposed to extracting semantics from a database schema. While database schema may contain semantics, it has been illustrated that the main focus of the database schema is the structure of data. As such, the semantics in a database schema are implicit and not maintainable. On the other hand, the main focus of ontologies is the semantics. As such, ontologies provide semantics that are explicit, maintainable, and up to date. Given the heterogeneous nature of open environment, a data integration system in open environment cannot rely on database schemas as primary sources for the semantics. It is also shown in Table 13 that the proposed model addresses the distributed nature of open environment through the use of a Mediated P2P architecture. The architecture proposed in (Xue 2010) is a mediated architecture. Mediated architecture is inherently centralized and does not address the distributed nature of open environment. (Y. D. Wang 2009), however, propose the use of a distributed architecture. It is worth mentioning that, with the absence of a centralized control, the system presented in (Y. D. Wang 2009) will
require each information system to act as a DIS on its own. This is too much to expect from every single information system in open environment.
When it comes to matching between ontologies of various information systems both (Y. D. Wang 2009) and (Xue 2010) propose the use of elementary ontology matching algorithms. The proposed model however proposes the use of a structural ontology matching algorithm. It has been demonstrated that the elementary ontology matching algorithms are not expected to yield accurate results. This is because elements of ontologies inherit semantics from their parents in the taxonomical structure. Elementary matching algorithms take concepts out of their context. As such, all the semantics that concepts inherit from their parents are not utilized by elementary matching algorithm. This is the main reason why elementary matching algorithms do not yield accurate results.
Because the two frameworks proposed in (Y. D. Wang 2009) and (Xue 2010) are based on an extensional reduction model, the two models use extensional equivalence. The extensional equivalence considers two predicates to be equivalent if they share an equivalent set of parameters. This is acceptable when dealing with a system that is extensional in nature. Data integration systems in open environment, however, are intensional in nature. As such, the proposed model employs intensional equivalence. According to the proposed model; two queries are considered to be intensionally equivalent if they are expressed in terms of equivalent intensional entities.
The two frameworks proposed in (Y. D. Wang 2009) and (Xue 2010) are based on extensional reduction model. This reflects on their description of a conceptualization. It has been demonstrated in Chapter 3 that the extensional reduction model is inadequate for describing a conceptualization. This is because it reduces the intensional matters to extensional entities. The proposed model however adopts a non reductionist approach that is based on the theory of PRP (Bealer 1979). The result is an intensional description of conceptualization. The proposed description is consistent with the view in (Guarino, Oberle, and Staab 2009). According to (Guarino, Oberle, and Staab 2009),
conceptualization is about meanings. As such, a conceptualization should not change unless meanings change.
And finally, it is shown in Table 13 that, the system proposed in (Y. D. Wang 2009) does not address the representation of ontology. On the other hand, the framework presented in (Xue 2010) proposes the use of a frame-based language for representing an ontology. It has been shown in section 2.2.4.2 that frame-based languages have limited expressive power and their semantics are not precisely defined. The proposed model, however, uses intensional logic to represent ontologies. Not only does the intensional logic have clear semantics, but also, intensional logic employs singular terms to express the properties of the concepts, and the relations between concepts.
Table 13: Comparison of Data Integration Framework in Open Environment
Factor Proposed Framework (Y. D. Wang 2009) (Xue 2010)
Model Intensional Extensional Reduction Extensional Reduction
Dynamic Nature Addressed with the
intensional model
Not addressed Not Addressed
Source of Semantics Ontology Ontological View (Extensional Reduction) D.B Schema
Architecture Mediated P2P Distributed (Web Services and Agents) Mediated
Mapping Structural and
Semantic-based
Elementary Elementary and
syntactical-based
Equivalence Intensional Extensional Extensional
Conceptualization Intensional Extensional Reduction Extensional Reduction