Robust Hate Speech Detection in Social Media: A Cross-Dataset Empirical Evaluation