. 2022 Oct 20;10(10):e41136. doi: 10.2196/41136

Table 9.

The comparison of mean F₁ scores of using different models to extract syntactic dependency features in the relation extraction of Bacteria-Biotope task (BB-rel) and the binary relation extraction of plant seed development task (SeeDev-binary).

Syntactic dependency feature	BB-rel			SeeDev-binary
	Minimum^a	Maximum^b	Mean (SD)	Minimum^a	Maximum^b	Mean (SD)
No-use	58.51	63.70	60.85 (1.65)	32.89	36.53	34.60 (1.16)
LSTM^c	59.93	65.16	62.88 (1.66)	34.55 ^d	37.90	36.06 (1.07)
CNN^e	60.68	65.56	63.13 (1.55)	34.47	38.04	36.37 (1.13)

^aThe lowest F₁-scores of 10 different random initializations.

^bThe highest F₁-scores of 10 different random initializations.

^cLSTM: long short-term memory network.

^dThe maximum results are shown in italics.

^eCNN: convolutional neural network.