There are approaches computing the semantic
relatedness of concepts with similarities to our
approach (e.g. Gabrilovich & Markovitch, 2007;
Kulkarni & Caragea, 2009). Gabrilovich and
Markovitch (2007) utilise the vast amount of
organized human knowledge that is available in
knowledge repositories like Wikipedia and Open
Directory Project (ODP). Each node in ODP is
treated as a concept. A textual object is created for
each node consisting of concatenated Web
documents (listed for each node by ODP) and their
textual descriptions. The concepts are represented
as attribute vectors. A document is divided into
non-overlapping segments called contexts where
each context is classified into one or several
concepts. An ambiguous concept will be part of
several domains, which is partly resolved by
categorising them. In the case of hierarchies, a
parent node will typically consist of both the child
concepts and a textual description.