Merging of Distributed Topic Maps based on the Subject Identity Measure (SIM) Approach
Lutz Maicher, Hans Friedrich Witschel · 2004
The central theoretical criteria of Topic Maps “One Topic for one Subject” leads to serious problems if two distributed Topic Maps are merged: according to existing standards, two Topics will only be merged if the description of their Subject (i.e. their so-called Subject Identifier or Subject Locator) is exactly identical. On the other hand – from a philosophical point of view – two topics should be merged if they describe the same Subject, i.e. if they are intended to refer to the same thing or idea. In distributed environments, however, Topic Map authors are not always able to use a common vocabulary: in these cases they will fail to use identical Subject Identifiers/Locators even if they intend to describe the same Subject. Therefore, we propose the SIM (Subject Identity Measure) approach which is based on a statistics using different Topic characteristics. This approach is on the one hand independent of the languages used and on the other hand of the structure in these Topic Maps. The SIM describes how closely related the Subjects of two distributed Topics are, even if the authors didn’t use a common vocabulary. Our algorithm uses as much information as possible in order to support users in decisions about which Topics to merge. If the SIM exceeds a given threshold, this indicates that two Topics describe the same Subject and therefore merging of these Topics will be recommended after a filtering process. Because Topic Maps are translatable into RDF and OWL the reuse of the SIM in Semantic Web applications should be enforced.