Databases Reference
In-Depth Information
class). Interestingly, the lattices are found automatically [
Bouzeghoub et al. 2000
],
but the merging is largely manual. To determine how to merge a node in the lattice,
they distinguish four cases:
There is one class at the node:
In this case, the answer is found automatically;
there are no conflicts and the class at that node is added to the resulting merged
ontology.
There are two or more classes at the node:
In this case, the user is asked what
should be done.
There are no classes at a nonroot node:
Here, the user must decide whether to
add a class or not.
There are no classes at a root node:
In this final case, the user must decide
whether to add a new top level class to the resulting ontology.
As seen from the description, FCA Merge makes no attempt to resolve any
conflicts.
5.4
Ontology Merging Analysis
Each of the three systems, SMART, Chimæra, and FCA Merge, takes a very differ-
ent approach to merging ontologies. Unlike database research (i.e., view integration
and data integration), all three systems view the problem of merge to intrinsically
require user intervention. SMART takes the most automatic approach of the three by
merging some concepts from different ontologies without requiring any user inter-
vention, but even this is limited: the user still must guide the system whenever the
names of the classes that match are too different. Even if the names are linguisti-
cally similar, there is little that SMART can do automatically other than point the
user at any potential conflicts unless the choice is clear from the preferred ontology.
Chimæra provides very little automatic support; it focuses the user's attention on
possibly related classes but has no conflict resolution support. FCA Merge provides
amazing support for automatically matching the classes in the ontologies including
doing some very sophisticated linguistic checks, but provides very little support for
automatically merging classes in the ontology if any sort of conflict exists.
Together, these solutions define an overall compendium of interesting and useful
features for ontology merging. SMART provides the notion of a preferred ontol-
ogy that can help the system to work automatically. They also suggest the process
of maintaining a list for the user of both where the user
must
perform an action
and where the user
should
perform an action with the Conflict and ToDo lists. The
Chimæra system offers good guidelines on what interactions must be available to
merge ontologies. Finally, FCA Merge introduces the notion of additional nodes
that are not present in either original ontology but may make the structure of the
resulting ontology more sensible.
