To evaluate the effect of the experiments we
needed a baseline (denoted as Bl in Table 2). The
baseline was conducted twice, at the beginning and
at the end of the experiments, to discount the effect
of uncontrollable external changes (e.g., change of
ranking by the search engine provider). To
measure this drifting effect, Definition 2 was used.
For the domain identification component (Step 7),
we selected to use parent entities for comparison
since it must be compared with at least one
neighbouring entity (see Step 7 in Section 2 for
details). The experiments were done over a period
of one week.