Table 9.

Comparison of different variants of our method on Event Dataset. The best performance indicators of all the taxonomies are marked as bold.

	ori.	cri. 1	cri. 2	cri. 3	cri. 4	cri. 5	$\overline{\mathit{\xi }}$ ± s
with all components	0.66	0.85	0.88	0.73	1.06	1.42	0.93 ± 0.25
No log-ratio loss	0.35	0.50	0.54	0.39	0.64	1.14	0.59 ± 0.26
No feature learning	0.29	0.39	0.43	0.33	0.47	0.92	0.47 ± 0.21
No active selection	0.63	0.79	0.85	0.68	0.93	1.31	0.87 ± 0.22
No verification	0.44	0.45	0.46	0.44	0.48	0.52	0.46 ± 0.03
No weighted blending	0.31	0.44	0.63	0.33	0.65	1.14	0.58 ± 0.28
No CRF	0.61	0.79	0.85	0.66	1.02	1.34	0.88 ± 0.25
No verification stopping	0.67	0.83	0.83	0.72	1.06	1.30	0.90 ± 0.22