Information extraction from structured documents using k-testable tree automaton inference