We have tested this by developing a small prototype that used thecorpora of labeled documents with some different learning algorithmsto see if the results would be satisfactory. We conclude that while thesystem would indeed ...