Semantic interoperability between relational databases in large-scale environments

As the number of relational database systems (RDBSs) on the web increases and evolves, the demands for semantic interoperability between these RDBSs increase. In such large-scale environments, exchanging and combining data without semantics will ultimately fail because participating RDBSs are designed and built independently so terms and their associated meanings defined and used in the individual RDBSs are likely to differ. This dissertation proposes a general purpose framework that uses the Semantic Web (SW) infrastructure for enabling semantic interoperability between autonomous RDBSs on the web. In addition to using a common data model and syntax to describe web resources, the proposed framework also utilizes the host ontology and the Relational Database Ontology (RDBO) defined in this dissertation to describe the host systems and RDBSs on the web. Using these two ontologies guarantees that systems and applications on the web can work with data instances that conform to a set of known vocabularies and structures. However, even for domain experts, constructing database ontologies (i.e., instances of the RDBO) describing the underlying RDBSs is time-consuming and error-prone. This dissertation describes a formal algorithm for generating database ontologies automatically and dynamically from RDBSs while preserving the underlying structural constraints. Implementation and experimentation of the algorithms are also discussed.; Semantic mappings between the database ontologies and/or the existing ontologies: (i) enable the RDBSs described by the database ontologies to interoperate semantically with each other; and (ii) enables the RDBSs described by the database ontologies to interoperate with the existing applications and data providers described by the existing ontologies. For example, given a domain ontology and a set of semantic mappings between the database ontologies (or ontologies in general), some semantic mappings can be decided and validated automatically by a reasoner engine that they are more semantically closer to the semantic information provided in the domain ontology than others. Without the domain ontology, no such claims can be made. This dissertation also modifies the Pellet reasoner engine and uses it to automatically validate and rank semantic mappings between the ontologies against their domain ontologies. Furthermore, this dissertation presents a semantic mapping scheme that supports semantic mappings between the ontologies with translations.