Name: GUSTAVO LUDOVICO GUIDONI
Publication date: 29/03/2023
Advisor:
Name |
Role |
|---|---|
| JOÃO PAULO ANDRADE ALMEIDA | Advisor * |
Examining board:
| Name |
Role |
|---|---|
| JOÃO PAULO ANDRADE ALMEIDA | Advisor * |
| MONALESSA PERINI BARCELLOS | Internal Examiner * |
| VÍTOR ESTÊVÃO SILVA SOUZA | Internal Examiner * |
Summary: Despite the relevant contributions of ontology-based conceptual modeling and the widespread
use of relational schemas, the combination of these two technologies has not yet received due
attention. Among the conceptual modeling technologies, OntoUML stands out as a language
to describe a domain of interest, having as its main niche the formulation and propagation of
knowledge. Conceptual models produced with OntoUML can be seen as a starting point for
other artifacts, such as relational schemas in a database realization. To produce a relational
schema from the conceptual model in an automated way, it is necessary to bridge the gap
between a series of constructs. The current literature provides some object-relational transfor-
mation approaches that could, in principle, be applied to ontology-driven conceptual models,
such as those produced in OntoUML. However, there are important constructs that are not
covered by such approaches that must be addressed. Most of the existing object-relational
transformation approaches fail to support conceptual models that: (i) include overlapping or
incomplete generalizations; (ii) support dynamic classication; (iii) have multiple inheritance;
and (iv) have orthogonal hierarchies. This is because many of the approaches discussed in the
literature assume primitives underlying object-oriented programming languages (instead of
conceptual modeling languages). To solve this gap, this work aims to understand the forces that
govern classical strategies for transforming class hierarchies into relational schemas, while rais-
ing some ontological meta-properties that characterize the classes in these models (like sortality
and rigidity). The information obtained is used to guide the transformation of the conceptual
model into a relational schema in order to avoid some problems in existing approaches, leading
to the novel one table per kind strategy. In addition to automating relational schema generation,
we also propose an automated ontology-based data access mapping for the resulting relational
schema, in order to provide access data in terms of the original conceptual model, and hence
queries can be written at a high level of abstraction (in SPARQL), independently of the transfor-
mation strategy selected. Further, we forward engineer additional constraints along with the
transformed schema (ultimately implemented as triggers) to guarantee that the semantics of the
source model is respected. The proposed approach is contrasted with dominant transformation
approaches in the literature from the perspectives of: (i) the supported conceptual modeling
primitives; (ii) size of the resulting schema; (iii) query answering performance; and (iv) usability
of the resulting schema, for which an empirical study is reported.
