UnQovering Stereotyping Biases via Underspecified Questions - EMNLP 2020 (Findings) - View it on GitHub