Towards Robust and Privacy-preserving Text Representations
This addresses privacy and fairness issues in NLP for users and developers, though it is incremental as it builds on existing representation learning methods.
The paper tackles the problem of author characteristics in text causing privacy risks and biased model performance by proposing a training approach to obscure these attributes, resulting in increased privacy and more robust models across different evaluation conditions.
Written text often provides sufficient clues to identify the author, their gender, age, and other important attributes. Consequently, the authorship of training and evaluation corpora can have unforeseen impacts, including differing model performance for different user groups, as well as privacy implications. In this paper, we propose an approach to explicitly obscure important author characteristics at training time, such that representations learned are invariant to these attributes. Evaluating on two tasks, we show that this leads to increased privacy in the learned representations, as well as more robust models to varying evaluation conditions, including out-of-domain corpora.