Ontology Alignment - Karlsruhe

We introduce the process of ontology alignment we use for the experiments here. The general process and an efficient algorithm for it have been presented in [ES04b,ES04a]. The presented approach is only capable of extracting semantical identity. Subsumption or complex alignments are currently not supported. To find out whether two entities can be aligned we rely on rules which indicate support for an alignment. A rule consists of two main parts: an entity feature, and a comparison function. The result is a value between 0 and 1. An entity is described through features which are represented in the ontology. A feature can be as simple as the label of entity. It can include intensional structural elements such as superor sub-concepts for concepts, or domain and range for relations. Further we use extensional elements: a concept is described through its instances. Instance features can be instantiated attributes. According to our assumption we restrict our focus on these simple features, where simple means that only directly connected relations are respected, i.e. we do not consider the relations of a sub-concept of a super-concept of a concept. The following example will clarify the used features.