Databases Reference
In-Depth Information
5
Ontology Evolution
Gruber
(
1993
) characterizes an ontology as the explicit specification of a concep-
tualization of domain. While there are different kinds of ontologies, they typically
provide a shared/controlled vocabulary that is used to model a domain of interest
using concepts with properties and relationships between concepts. In the recent
past, such ontologies have been increasingly used in different domains to seman-
tically describe objects and to support data integration applications. For example,
there are a growing number of life science ontologies, e.g., the ontologies managed
in the open biomedical ontologies (OBO) Foundry (
Smith et al. 2007
). The exist-
ing ontologies are not static but are frequently evolved to incorporate the newest
knowledge of a domain or to adapt to changing application requirements.
There are several differences between ontologies and relational schemas that
influence their evolution:
Ontologies are conceptually more abstract models than database schemas and
come in different variations ranging from controlled vocabularies and thesauri
over is-a hierarchies/taxonomies and directed a-cyclic graphs (DAG) to frame-
based and formal representations (
Lassila and McGuinness 2001
). For instance,
ontology languages such as RDF or OWL allow the specification of concept
hierarchies with multiple inheritance, cardinality constraints, inverse or transitive
properties, and disjoint classes. The kind and expressiveness of ontologies deter-
mine the kind of changes that should be supported for ontology evolution. For
instance,
Noy and Klein
(
2004
) propose a set of 22 simple and complex ontol-
ogy change operations such as concept creation, reclassification of a concept, or
merge/split of concepts.
The role of instances differs between ontologies and relational schemas. For
example, many ontologies include instances but do not clearly separate them
from other parts of the ontologies such as concepts and relationships. In other
cases, instances are described by ontologies but are maintained outside the ontol-
ogy within separate data sources. These differences impact update propagation
of ontology changes since the separately maintained instances may not be under
the control of the ontology editors.
In contrast to database schemas, the development and evolution of ontologies
is often a collaborative and decentralized process. Furthermore, new ontologies
often reuse existing ones, i.e., an ontology engineer uses a common ontology as
the basis for domain-specific extensions. These aspects lead to new synchroniza-
tion requirements for ontology changes. Furthermore, ontologies serving a whole
domain likely introduce many usage dependencies, although ontology providers
usually do not know which applications/users utilize their ontology. Supporting
different ontology versions is a main approach to provide stability for ontology
applications. For example, there are daily new versions for the popular Gene
Ontology.
Despite these differences, it is easy to see that the schema evolution requirements
introduced in Sect.
2
also apply to ontology evolution, in particular support for a rich
