Towards Understanding Gender Bias in Relation Extraction
This addresses social bias in automated knowledge base construction, an incremental step as it introduces the first evaluation framework for bias in NRE.
The paper tackled the problem of gender bias in Neural Relation Extraction (NRE) systems by creating WikiGenderBias, a dataset to evaluate bias, and found that these systems exhibit gender-biased predictions, with analysis showing effects of techniques like name anonymization and debiasing.
Recent developments in Neural Relation Extraction (NRE) have made significant strides towards Automated Knowledge Base Construction (AKBC). While much attention has been dedicated towards improvements in accuracy, there have been no attempts in the literature to our knowledge to evaluate social biases in NRE systems. We create WikiGenderBias, a distantly supervised dataset with a human annotated test set. WikiGenderBias has sentences specifically curated to analyze gender bias in relation extraction systems. We use WikiGenderBias to evaluate systems for bias and find that NRE systems exhibit gender biased predictions and lay groundwork for future evaluation of bias in NRE. We also analyze how name anonymization, hard debiasing for word embeddings, and counterfactual data augmentation affect gender bias in predictions and performance.