| Abstract |
The lack of a central naming authority is a key characteristic of the
World Wide Web and the Semantic Web alike. This means that data
publishers are free to introduce IRIs that refer to the same instance
or concept. In order to still be able to meaningfully interchange
data on the web, it is necessary to introduce links between these
IRIs. `owl:sameAs' is by far the most common property for this sort
of interlinking.
I will present https://sameAs.cc, a resource that computes the
equivalence closure for 558M explicit `owl:sameAs' statements
obtained from the web. The closure consists of 35B implicit
statements. As is to be expected, some of these 35B implicit
are incorrect, e.g., stating that The Netherlands is the same as
Germany. There are also more subtle errors, e.g., equating the person
Barack Obama to the 44th president of the U.S. I will present an
approach for automatically detecting identity communities, i.e.,
subgraphs of identity graphs that have relatively few errors and that
therefore are a better starting point for interlinking. |