Figure 5.

(A) Receiver operating characteristics and (B) precision-recall curves for all machine learning methods (solid lines) and the naïve word-matching method (dotted line) in the training set (Erlangen-A).