Adversarial Examples for Natural Language Classification Problems