Table 2.

Results across all users for BioCreAtIvE Task 2.2.

User, Run	# results	"perfect"	"generally"
4, 1	661	78 (11.80%)	49 (7.41%)
7, 1	153	1 (0.65%)	1 (0.65%)
7, 2	384	19 (4.95%) [1]	9 (2.34%) [1]
7, 3	263	2 (0.76%)	10 (3.80%)
9, 1	28	9 (32.14%)	3 (10.71%)
9, 2	41	14 (34.15%)	1 (2.44%)
9, 3	41	14 (34.15%)	1 (2.44%)
10, 1	120	35 (29.17%)	8 (6.67%)
10, 2	86	24 (27.91%)	6 (6.98%)
10, 3	116	37 (31.90%)	11 (9.48%)
15, 1	502	3 (0.60%)	8 (1.59%)
15, 2	485	16 (3.30%)	26 (5.36%)
17, 1	247	52 (21.05%) [1]	23 (9.31%) [0]
17, 2	55	1 (1.82%)	0 (0.00%)
17, 3	99	1 (1.01%)	1 (1.01%)
20, 1	673	20 (2.97%)	30 (4.46%)
20, 2	672	38 (5.65%)	26 (3.87%)
20, 3	673	58 (8.62%)	27 (4.01%)

Evaluation results on the evidence text selected for Task 2.2. See legend for Table 1. In this task, evaluation of the GO node reference was done with respect to the predicted GO annotation provided by the system.