An Overview of Biomolecular Event Extraction from Scientific Documents

Table 7

Negation and speculation detection performance comparison. BioNLP shared task comparison results in recall/precision/F-score (%) on the test set for Task 3 (negation/speculation detection). (A) abstracts only and (F) full papers only. Data extracted from BioNLP-ST 2009, BioNLP-ST 2011, and BioNLP-ST 2013 overviews [51, 52, 109].

Year	System		Negation	Speculation	Total
2009	ConcordU09; Kilicoglu and Bergler [84]	(A)	14.98/50.75/23.13	16.83/50.72/25.27	15.86/50.74/24.17
2011	UTurku; Björne et al. [64, 77]	(A); (F)	22.03/49.02/30.40; 25.76/48.28/33.59	19.23/38.46/25.64; 15.00/23.08/18.18	20.69/43.69/28.08; 19.28/30.85/23.73
	ConcordU11; Kilicoglu and Bergler [93]	(A); (F)	18.06/46.59/26.03; 21.21/38.24/27.29	23.08/40.00/29.27; 17.00/34.69/22.82	20.46/42.79/27.68; 18.67/36.14/24.63
2013	TEES-2.1; Björne and Salakoski [97]	(F)	21.68/36.84/27.30	18.46/33.96/23.92	19.53/35.59/25.22
	EVEX; Hakala et al. [91]	(F)	20.98/38.03/27.04	18.46/32.73/23.61	19.82/34.41/25.15