Generalization of our NLP pipeline on epileptologist, neurologist, and non-neurologist notes from the University of Pennsylvania, and on epileptologist notes from the University of Michigan, measured as agreement between algorithm predictions and human-annotated ground truth.