Applied Semantic Web Technologies

226  ◾  Dae-Ki Kang

to construct a taxonomy



T

by starting with the primitive attributes in à as the

leaves of



T

and recursively adding nodes to



T one at a time by merging two exist-

ing nodes.

Let DM(P(x)||Q(x)) denote a measure of pair-wise divergence between two

probability distributions P and Q of the random variable x. We use a pair-wise

measure of divergence between the distributions of the class labels associated

with the corresponding Boolean attributes as a measure of dissimilarity. e

lower the divergence between the class distribution of two attributes, the greater

is their similarity. e choice of this measure is motivate ...