46
is indeed an instance of the class
City.
Using a set of patterns
R expressing relation
r, we formulate the following
acceptance function:
accept
c
s
(
t) =
½
true
if ∑
S∈R
h(
S, c
s
,t)
≥ n
false otherwise
where
h(
S, c
s
,t) is the number of hits for query with pattern
S combined with
term
t and the plural form of the name of class
c
s
. The threshold
n has to be
chosen beforehand. We can do so, by calculating the sum of
hits for queries with
known instances of the class. Based on these figures, a threshold can be chosen
e.g. the minimum of these sums. When the instances in the initial ontology are
well-known, the sum of hits for these instances can be expected to be large. Hence,
setting a threshold based on such instances will lead to a threshold (and acceptance
function) that will filter out correct, but less well-known, instances.
When we use such an acceptance function, we can allow ourselves to formulate
less strict recognition rules. That is, false instances that are at first accepted, are
still rejected as an instance by the use of the acceptance function.
As an alternative, a term
t can be checked using
Google’s define functionality.
If the name of
c
Dostları ilə paylaş: