Adversarial Attacks and Dimensionality in Text Classifiers