Table 3.

Validation set performances of the LSTM models with mean pooling word embedding trained on medical corpus. The hyper-parameters used by each model are the same as specified in Table 2. The numbers in parentheses are the standard deviations of the corresponding metrics across the hyper-parameter grids.

System	Problem-Treatment (TrP) Relations			Problem-Test (TeP) Relations			Problem-Problem (PP) Relations
	R	P	F	R	P	F	R	P	F
Segment LSTM mean	0.788 (0.019)	0.767 (0.022)	0.777 (0.009)	0.826 (0.019)	0.843 (0.014)	0.834 (0.008)	0.784 (0.029)	0.784 (0.032)	0.784 (0.017)
Sentence LSTM mean	0.783 (0.033)	0.770 (0.016)	0.776 (0.011)	0.846 (0.019)	0.806 (0.022)	0.825 (0.008)	0.853 (0.010)	0.818 (0.005)	0.835 (0.004)