Table 4.

The overall performance of the hybrid system. Entity-Strict level evaluation results, which are the primary results used by the N-GRID challenge [15], are shown in bold face.

PHI category set	Evaluation level	Precision	Recall	F1-measure
N-GRID categories	Token	95.23	90.15	92.62
	Entity-Strict	93.39	88.23	90.74
	Entity-Relaxed	93.5	88.33	90.84
	Binary-Token	97.84	92.62	95.16
	Binary-Strict	95.66	90.38	92.94
HIPAA categories	Token level	96.3	92.24	94.23
	Entity-Strict	94.56	90.47	92.47
	Entity-Relaxed	94.67	90.57	92.58
	Binary-Token	97.18	93.09	95.09
	Binary-Strict	95.34	91.21	93.23