Sampling cases with differential probability, so that classes that occur rarely in the population occur more often in the training data. Does not address the problems in unbalanced classes.
unbalanced-classes do pose problems, but contrary to common misunderstandings, these are merely due to low sample size (high variance of predictors), not the unbalancedness per se. As such, oversampling will not help.
See Are unbalanced datasets problematic, and (how) does oversampling (purport to) help? and links there.
