Automatic classification of relevant/irrelevant documents from a larger set.