Reliability Testing for Natural Language Processing Systems
This addresses the need for accountability in deploying NLP systems, particularly for fairness across demographics, but is incremental as it builds on existing adversarial attack methods.
The paper tackles the problem of ensuring NLP systems are fair, robust, and transparent by proposing reliability testing, which reframes adversarial attacks to develop targeted tests for diverse and noisy environments.
Questions of fairness, robustness, and transparency are paramount to address before deploying NLP systems. Central to these concerns is the question of reliability: Can NLP systems reliably treat different demographics fairly and function correctly in diverse and noisy environments? To address this, we argue for the need for reliability testing and contextualize it among existing work on improving accountability. We show how adversarial attacks can be reframed for this goal, via a framework for developing reliability tests. We argue that reliability testing -- with an emphasis on interdisciplinary collaboration -- will enable rigorous and targeted testing, and aid in the enactment and enforcement of industry standards.