Parallel sequence tagging for concept recognition

BMC Bioinformatics

Table 3 Results for our current BioBERT system, best system reported in the shared-task paper [46], and the official baseline

Annot. set	System	Proper		Extended
Annot. set	System	SER	F-score	SER	F-score
CHEBI	Baseline	0.44	0.72	0.29	0.80
	Shared-task	0.3388	0.7700	0.2571	0.8209
	Current	0.2492	0.8528	0.2289	0.8459
CL	Baseline	0.53	0.61	0.33	0.73
	Shared-task	0.4862	0.6657	0.3361	0.7484
	Current	0.4013	0.7526	0.2777	0.7926
GO_BP	Baseline	0.39	0.72	0.29	0.79
	Shared-task	0.3047	0.8037	0.2786	0.8138
	Current	0.2587	0.8297	0.2015	0.8506
GO_CC	Baseline	0.44	0.71	0.20	0.88
	Shared-task	0.3788	0.7645	0.1678	0.8936
	Current	0.2817	0.8219	0.1486	0.9073
GO_MF	Baseline	0.07	0.95	0.45	0.66
	Shared-task	0.0319	0.9838	0.3881	0.7438
	Current	0.0149	0.9904	0.4135	0.7139
MOP	Baseline	0.43	0.75	0.36	0.79
	Shared-task	0.2684	0.8705	0.3080	0.8437
	Current	0.1567	0.9188	0.1713	0.9082
NCBITaxon	Baseline	0.07	0.96	0.07	0.96
	Shared-task	0.0537	0.9694	0.0466	0.9722
	Current	0.0436	0.9744	0.0460	0.9704
PR	Baseline	0.69	0.48	0.62	0.52
	Shared-task	0.3052	0.8026	0.3030	0.8011
	Current	0.3068	0.8041	0.3130	0.7951
SO	Baseline	0.21	0.86	0.18	0.89
	Shared-task	0.1593	0.9027	0.1230	0.9187
	Current	0.1206	0.9223	0.0899	0.9419
UBERON	Baseline	0.41	0.70	0.36	0.75
	Shared-task	0.3752	0.7488	0.3371	0.7714
	Current	0.2790	0.8177	0.2537	0.8315

In case of the shared-task systems, the results were selected independently for SER and F-score, i. e. the two scores for a given annotation set do not necessarily come from the same system. For the baseline and the current BioBERT system, however, only one system was evaluated for each annotation set

ISSN: 1471-2105