The main aim for this dataset is to cover a wide variety of social biases that are implied in text, both subtle and overt, and make the biases representative of real world discrimination that people experience RWJF 2017.