Text this: A Robust Method to Protect Text Classification Models against Adversarial Attacks