Gender, race and social biases have recently been detected as evident examples of unfairness in applications of Natural Language Processing. A key path towards fairness is to understand, analyse and interpret our data and algorithms. Recent studies have shown that the human-generated data used in training is an apparent factor of getting biases. In addition, current algorithms have also been proven to amplify biases from data. To further address these concerns, in this paper, we study how an state-of-the-art recurrent neural language model behaves when trained on data, which under-represents females, using pre-trained standard and debiased word embeddings. Results show that language models inherit higher bias when trained on unbalanced data when using pre-trained embeddings, in comparison with using embeddings trained within the task. Moreover, results show that, on the same data, language models inherit lower bias when using debiased pre-trained emdeddings, compared to using standard pre-trained embeddings.
View on arXiv