A Dataset for Tackling Gender Bias in Text

Gender bias is found in personal conversations, in the media, in historical writings, popular culture, in the labor force, household responsibilities and now in machines (Fiebert and Meyer, 1997; Kingdon, 2005). Gender bias occurs in machine learning models when they are trained with data that contains human-like biases (Haussler, 1988). Current research is focused on detecting and correcting for gender bias in existing machine learning models, such as word embeddings (Zhao et al., 2018: Bolukbasi et al., 2016), coreference resolution (Zhao et al., 2018) and visual recognition tasks involving language (captioning) (Zhao et al., 2017). Rather than removing gender bias in current machine learning models, we are tackling the issue at its root and creating a gender bias dataset with which to train a machine learning model. Enabling a model to learn gender bias would allow for gender bias detection and possibly correction in text.