... they rely on specific textual features that are not universal to all text types. More recent approaches to automatic text deidentification have used deep-learning techniques, typically transformer-based language models. A transformer model is a neural network architecture widely used in NLP (Vaswani...