Table 1.

Candidate-mention scores of information extraction applications trained with data programming using generative models with no dependency structure (Independent) and learned dependency structure (Structure).

Application	Independent			Structure			F1 Diff.	# LFs	# Cor.	% Corr.
	P	R	F1	P	R	F1
Disease Tagging	60.4	73.3	66.3	68.0	69.8	68.9	2.6	233	315	1.17%
Chemical-Disease	45.1	69.2	54.6	46.8	69.0	55.9	1.3	33	21	3.98%
Device Polarity	78.9	99.6	88.1	80.5	98.6	88.7	0.6	12	32	48.49%