Our approach to the construction of FVs is
based on a non-supervised solution that is
applicable to any ontology and text corpus as long
as there is some correlation between them.
However, the approach is not dependent on a
collection of only relevant documents. On the
contrary, the main advantage of the approach is
that a diverse corpus, like the Web can be used.
Our approach is capable of disambiguating word
sense by utilising the relationships between the
entities. Nevertheless, the FV quality will be
highly dependent on both the quality of the
ontology and correlation of terminologies in the
ontology and the text corpus.