Internal semantics in syntactic networks that learn
A network of appropriately connected syntactical symbols—which can learn by deriving consequences from inputs—possesses internal semantics and can be said to understand as human understanding seems to result from the same kind of internal semantics.
William Rapaport, 1988.