Data driven selection of positive documents (remove spam documents) without NLP preprocessing - View it on GitHub