Mapping Ontologies by Utilising Their Semantic Structure

Background

Borrowed from philosophy, ontology refers to a systematic account of what can exist or ‘be’ in the world. In the fields of artificial intelligence and knowledge representation, ontology refers to the construction of knowledge models that specify a set of concepts, their attributes, and the relationships between them. Ontologies are defined as “explicit conceptualisation(s) of a domain” (Gruber, 1993), and are seen as a key to realise the vision of the Semantic Web.

Ontology, as an important technique to represent knowledge and information, allows to incorporate semantics into data to drastically enhance information exchange. The Semantic Web (Berners-Lee, Hendler & Lassila, 2001) is as a universal medium for data, information, and knowledge exchange. It suggests to annotate web resources with machine-processable metadata. With the rapid development of the Semantic Web, it is likely that the number of ontologies used will strongly increase over the next few years. By themselves, however, ontologies do not solve any interoperability problem. Ontology mapping (Ehrig, 2004) is, therefore, a key to exploit semantic interoperability of information and, thus, has been drawing great attention in the research community during recent years. This section introduces the basic concepts of information integration, ontologies, and ontology mapping.

Mismatches between ontologies are mainly caused by independent development of ontologies in different organisations. They become evident when trying to combine ontologies which describe partially overlapping domains. The mismatches between ontologies can broadly be distinguished into syntactic, semantic, and structural heterogeneity. Syntactic heterogeneity denotes differences in the language primitives used to specify ontologies, semantic heterogeneity denotes differences in the way domains are conceptualised and modelled, while structural heterogeneity denotes differences in information structuring.

There have been a number of previous works proposed so far on ontology mapping (Shvaiko, 2005, Noy, 2004, Sabou, 2006, Su, 2006). In (Madhavan, 2001), a hybrid similarity mapping algorithm has been introduced. The proposed measure integrates the linguistic and structural schema matching techniques. The matching is based primarily on schema element names, not considering their properties. LOM (Li, 2004) is a semi-automatic lexicon-based ontology-mapping tool that supports a human mapping engineer with a first-cut comparison of ontological terms between the ontologies to be mapped. It decomposes multi-word terms into their word constituents except that it does not perform direct mapping between the words. The procedure associates the WordNet synset index numbers of the constituent words with ontological term. The two terms which have the largest number of common synsets are recorded and presented to the user.