Table 3.

Results with KF/ELMo representations on the test set for PNER

Model	Strict Match			Overlap Match
	P	R	F1	P	R	F1
Baseline	0.749	0.816	0.781	0.800	0.871	0.834
+ linguistic features	0.817	0.764	0.789	0.868	0.812	0.839
+ linguistic features + KFs	0.776	0.845	0.809	0.820	0.893	0.855
+ linguistic features + ELMo	0.826	0.801	0.813	0.874	0.847	0.860
+ linguistic features + KFs + ELMo (ours)	0.815	0.812	0.814	0.873	0.869	0.871

Strict match criteria require that the predicted entity and the gold standard annotations have to match exactly at the byte offset; and overlap match criteria allows a match if the predicted entity overlaps with the gold annotation at all. The highest scores are highlighted in bold. We tune the hyper-parameters through the validation set and use the official evaluation script to assess the performance of the final chosen model on the test set