Reducing Gender Bias in Abusive Language Detection
This addresses the problem of model robustness for practical deployment in abusive language detection, though it is incremental as it applies existing mitigation techniques to this domain.
The paper tackled gender bias in abusive language detection models, which often misclassify non-abusive sentences containing identity words due to imbalanced training data, and achieved a 90-98% reduction in bias using methods like debiased embeddings and data augmentation.
Abusive language detection models tend to have a problem of being biased toward identity words of a certain group of people because of imbalanced training datasets. For example, "You are a good woman" was considered "sexist" when trained on an existing dataset. Such model bias is an obstacle for models to be robust enough for practical use. In this work, we measure gender biases on models trained with different abusive language datasets, while analyzing the effect of different pre-trained word embeddings and model architectures. We also experiment with three bias mitigation methods: (1) debiased word embeddings, (2) gender swap data augmentation, and (3) fine-tuning with a larger corpus. These methods can effectively reduce gender bias by 90-98% and can be extended to correct model bias in other scenarios.