Databases Reference
In-Depth Information
set of changes, expressive mappings, update propagation to instances and dependent
schemas/ontologies, versioning, and user-friendly tools.
For the rest of this section, we will describe representative approaches on
ontology evolution and how they meet the introduced requirements. Table
6.3
comparatively shows selected approaches that are discussed at the end of the section.
5.1
Research Approaches
The
Protege
system supports different kinds of collaborative ontology evolution
meeting varying functional requirements (
Noy et al. 2006
). First, ontologies can be
modified synchronously or asynchronously. Synchronous editing is performed on
a centrally stored ontology that can be modified concurrently by several develop-
ers. For asynchronous editing collaborators check out the latest ontology version,
change it offline, and merge their changes into a common version later on. Sec-
ond, ontologies may internally be versioned or not. Ontologies may so periodically
be archived with the possibility to roll back to a former version. Alternatively, all
changes are continuously directed to a single (the most recent) ontology version.
Third, ontology changes may be subject to the approval of designated curators to
resolve potential problems and maintain a high quality. Usually, such a curation is
performed before releasing a new version of an ontology. Finally, ontology changes
may be monitored (logged) or not.
The ontology evolution framework supports a rich set of simple and complex
changes that can be annotated (
Noy et al. 2006
). These changes are classified within
a change and annotation ontology (CHAO). Annotation includes the type of ontol-
ogy change, the class/property/instance that was changed, the user and date/time
when the change was performed. The two main approaches to specify changes are
supported: specification (and logging) of incremental change operations and the pro-
vision of a new ontology version. In the latter case, the Diff evolution mapping is
semi-automatically determined.
Protege uses the PROMPTDIFF algorithm (
Noy and Musen 2002
) to deter-
mine an evolution mapping between two input ontology versions. The two versions
V1 and V2 are compared using an iterative algorithm combining different heuris-
tic matchers (e.g., single unmatched sibling, unmatched inverse slots, or same
type/name) until no more changes are found. The found changes are presented in
a so-called difference table containing a set of tuples that interrelate elements of
V1 with elements of V2. Each tuple specifies a change operation (add, delete, split,
merge, and map) and its parameters.
The different kinds of ontology evolution are implemented in the Protege ontol-
ogy editor within two plugins: Change-management plugin and the PROMPT
plugin. The Change-management plugin can be used to access a list of changes,
allows users to add annotations, and enables to study the history of concepts, i.e.,
users can examine what modifications happened on a particular concept in the his-
tory. The PROMPT plugin implements the PROMPTDIFF algorithm and provides
