17-03-2014, 05:31 PM
Abstract
When we do ontology merging or aligning, it is
important to compare concepts which are likely to be
merging or aligning points in different ontologies. In
most cases, concepts with similar names are
considered to be promising candidates. However, it is
quite probable that concepts with lexically similar
names are not semantically similar or compatible. In
this paper, we presented a method to compare concepts
between two ontologies from four perspectives based
on the semantic distance which to capture the meaning
of a concept from its position in the containing
ontology as well as its defined attributes.
1. Introduction
Ontology has proven to be useful for knowledge
representation, information sharing and system
modeling. It is widely used in knowledge based
systems, semantic web, natural language processing,
and other applications.
An important task for ontology merging and
aligning is to detect similar concepts residing in two
ontologies, which provide likely points for merging
and aligning. A direct solution is to find concepts with
lexically similar names. However, it is probable that
two concepts with even the same name in two
ontologies may be semantically disparate. In order to
validate that concepts with similar names are really
semantically similar and compatible, we have to
compare concepts from deeper perspectives that reflect
concept meanings.
Previously, researchers in the field of database and
information system did some work that are in fact
concept comparison, though they may not use the
terminology concept, but type, class or object schema
instead. In mapping types, Lehmann and Cohn [4]
commanded that types (“eggs”) are supplied with
prototypes (“yolks”), and they used the intersections
between types as well as prototypes to determine the
reliability of mapping between types. Kashyap and
Sheth [3] adopted a context-based approach to evaluate
semantic and schematic similarities between database
objects. Weinstein and Birmingham [8] used three
kinds of measures they called: filter measures,
matching-based measures and probabilistic measures to
compute description compatibility of concepts in
differentiated ontologies.
Rather than measure semantic similarity directly, we
measure semantic distance instead. It doesn’t make
much difference since they are dual to each other. In
this paper, we proposed a method to compare the
semantic distance of concepts from four perspectives:
the super concepts set, the sub-concepts set, the
intention, and the extensionality. The difference of our
work from previous research is that we consider the
meaning of a concept in a specific ontology to be
captured both by the position of the concept in the
ontology as well as the attributes defined for the
concept.
In Section 2, we define our ontology model and give
an example of ontology. Section 3 gives the detail of
our concept comparison method. And Section 4 is the
conclusion.
When we do ontology merging or aligning, it is
important to compare concepts which are likely to be
merging or aligning points in different ontologies. In
most cases, concepts with similar names are
considered to be promising candidates. However, it is
quite probable that concepts with lexically similar
names are not semantically similar or compatible. In
this paper, we presented a method to compare concepts
between two ontologies from four perspectives based
on the semantic distance which to capture the meaning
of a concept from its position in the containing
ontology as well as its defined attributes.
1. Introduction
Ontology has proven to be useful for knowledge
representation, information sharing and system
modeling. It is widely used in knowledge based
systems, semantic web, natural language processing,
and other applications.
An important task for ontology merging and
aligning is to detect similar concepts residing in two
ontologies, which provide likely points for merging
and aligning. A direct solution is to find concepts with
lexically similar names. However, it is probable that
two concepts with even the same name in two
ontologies may be semantically disparate. In order to
validate that concepts with similar names are really
semantically similar and compatible, we have to
compare concepts from deeper perspectives that reflect
concept meanings.
Previously, researchers in the field of database and
information system did some work that are in fact
concept comparison, though they may not use the
terminology concept, but type, class or object schema
instead. In mapping types, Lehmann and Cohn [4]
commanded that types (“eggs”) are supplied with
prototypes (“yolks”), and they used the intersections
between types as well as prototypes to determine the
reliability of mapping between types. Kashyap and
Sheth [3] adopted a context-based approach to evaluate
semantic and schematic similarities between database
objects. Weinstein and Birmingham [8] used three
kinds of measures they called: filter measures,
matching-based measures and probabilistic measures to
compute description compatibility of concepts in
differentiated ontologies.
Rather than measure semantic similarity directly, we
measure semantic distance instead. It doesn’t make
much difference since they are dual to each other. In
this paper, we proposed a method to compare the
semantic distance of concepts from four perspectives:
the super concepts set, the sub-concepts set, the
intention, and the extensionality. The difference of our
work from previous research is that we consider the
meaning of a concept in a specific ontology to be
captured both by the position of the concept in the
ontology as well as the attributes defined for the
concept.
In Section 2, we define our ontology model and give
an example of ontology. Section 3 gives the detail of
our concept comparison method. And Section 4 is the
conclusion.