Automatic Evaluation of Search Ontologies ◾ 285
ontology individuals remained an acute challenge. In this experiment, we used the
corpus of reviews as a representative set of search queries. Note that we do not claim
that reviews are similar to actual search queries. We only claim that the way movies
and actors are named in naturally occurring reviews is a good indication of how
they will be named in search queries.
is exercise indicates how a textual corpus aligned with an ontology and
mature language technology (named-entity recognition and exible string simi-
larity) allows us to measure a complex property of an ontology. is evaluation
provides a score for the ontology and also indicates which specic named entities
are us ...