Power of Explanations: Towards automatic debiasing in hate speech detection