Learning and Evaluating a Differentially Private Pre-trained Language Model

Contextual language models have led to significantly better results on a plethora of language understanding tasks, especially when pre-trained on the same data as the downstream task. While this additional pre-training usually improves performance, it often leads to information leakage and therefore risks the privacy of individuals mentioned in the training data. One method to guarantee the privacy of such individuals is to train a differentially private model, but this usually comes at the expense of model performance. Moreover, it is hard to tell given a privacy parameter $\epsilon$ what was the effect on the trained representation and whether it maintained relevant information while improving privacy. To improve privacy and guide future practitioners and researchers, we demonstrate here how to train a differentially private pre-trained language model (i.e., BERT) with a privacy guarantee of $\epsilon=0.5$ with only a small degradation in performance. We experiment on a dataset of clinical notes with a model trained on an entity extraction (EE) task on and compare it to a similar model trained without differential privacy. Finally, we present a series of experiments showing how to interpret the differentially private representation and understand the information lost and maintained in this process.