On Bias and Fairness in NLP: How to have a fairer text classification?