Analyzing the fairness of natural language representations, such as sentence and document encodings, and proposing a method to improve fairness while maintaining classification accuracy.
