Towards Robust and Privacy-preserving Text Representations

Abstract

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.

arXiv commentAccepted to ACL 2018

References in corpus (1)

• Frustratingly Easy Domain Adaptation

Cited by in corpus (7)

• Generative Adversarial Networks: A Survey Towards Private and Secure Applications
• Adversarial Removal of Demographic Attributes from Text Data
• Benchmarking Differential Privacy and Federated Learning for BERT Models
• Anonymization for Skeleton Action Recognition
• MABR: Multilayer Adversarial Bias Removal Without Prior Bias Knowledge
• Privacy-preserving Neural Representations of Text
• Mitigating the Position Bias of Transformer Models in Passage Re-Ranking