Towards Automatic Generation of Shareable Synthetic Clinical Notes Using\n Neural Language Models

Abstract

Large-scale clinical data is invaluable to driving many computational\nscientific advances today. However, understandable concerns regarding patient\nprivacy hinder the open dissemination of such data and give rise to suboptimal\nsiloed research. De-identification methods attempt to address these concerns\nbut were shown to be susceptible to adversarial attacks. In this work, we focus\non the vast amounts of unstructured natural language data stored in clinical\nnotes and propose to automatically generate synthetic clinical notes that are\nmore amenable to sharing using generative models trained on real de-identified\nrecords. To evaluate the merit of such notes, we measure both their privacy\npreservation properties as well as utility in training clinical NLP models.\nExperiments using neural language models yield notes whose utility is close to\nthat of the real ones in some clinical NLP tasks, yet leave ample room for\nfuture improvements.\n

Ähnliche Arbeiten