The creation of attribute label in step 2 stems from the fact that ground facts of the form
p(C) can be rewritten as simple implications of the form label(x; C) ) p(x). Notice how
the attributes whose values are ? in a vector correspond exactly to those predicates that do
not appear in the premise of the corresponding FOL clause. The attribute corresponding
to q(x) has dierent usages. It functions as a conclusion during forward chaining and as a
target classication during inductive learning. In some cases, it can also be used as a goal.
To avoid unnecessary confusion, the attribute corresponding to q(x) is simply referred to
as the target-attribute. The values of the target-attribute are subsequently tagged with the
subscript T . The translation from FOL to AVL is currently performed manually