Quantifying Mention-Level Genderedness
Ananya, Nitya Parthasarthi and Sameer Singh
Appearing at NAACL 2019
Paper Source Code Demo (coming soon)
Language is gendered if the context surrounding a mention is suggestive of a particular binary gender for that mention. Detecting the different ways in which it occurs is an important task since gendered language can bias NLP models (such as for coreference resolution). This task is challenging since genderedness is often expressed in subtle ways. Existing approaches need considerable annotation efforts for each language, domain, and author, and often require handcrafted lexicons and features. Additionally, these approaches do not provide a quantifiable measure of how gendered the text is, nor are they applicable at the fine-grained mention level.
In this paper, we provide a method that automatically provides gender labels for mentions in the text, which leverages syntactic and se- mantic constructions. We further train a supervised classifier on corpora labeled using this method to predict the gender of any mention from its context and evaluate it on unseen text. The model confidence for a prediction can be used as a proxy to indicate the level of genderedness of the context. We test this gendered language detector on movie summaries, movie reviews, news articles, and fiction novels, achieving an AUC of up to 0.71, and observe that the model predictions agree with the human predictions collected for this task. We also perform a qualitative analysis, providing examples of detected gendered sentences from the aforementioned domains.