Visitor profiling is a reasonable server content filtering method, which allows
to provide content according to the visitor's preferences. The authors' goal is
developing a new approach to Internet portal visitors profiling and implementing
this method in form of a document web-server capable of forecasting the visitor's
interests.
The main notions in an ontology are called subjects, they are extracted from
the ontology based on certain criteria. Each ontology notion has a corresponding
vector, whose coordinates characterize the closeness of the notion to a subject. It
is possible to form a vector characterizing the closeness of a document to each
subject based on the vectors of notions mentioned in the document. This vector is
called a document profile. Each visitor has interests and therefore can be described
with a similar vector, called a visitor profile. However, in contrast to the
document vector, which is static, the visitor vector is adjusted each time the visitor
reads a new document. Therefore, the visitor profile changes as the visitor's
interests change. The visitor profile is compared with document profiles to find
out which documents are more interesting to the user at the moment. The goodness
is determined by the angle between the two vectors and their length. We can
get quite accurate estimation using simple scalar vector multiplication.
Keywords: Visitors profiling, Semantic Web, semantic networks, vector
model of information retrieval, ontology.
