Skip to main content

View full-text article in PMC

. 2023 Apr 14;4(4):100729. doi: 10.1016/j.patter.2023.100729

Table 10.

Comparison of PubMedBERT fine-tuned models, scispaCy, Stanza on BLURB named-entity recognition tasks (standard entity-level test F1 score)

	scispaCy			Stanza					PubMedBERT
	JNLPBA	bc5cdr	max	jnlpba	bc5cdr	ncbi-disease	bc4chemd	max	$B A S E$	$L A R G E$
BC5-chem	3.60	86.49	86.49	–	91.47	–	89.84	91.47	93.33^∗	93.23
BC5-disease	1.35	80.03	80.03	–	83.78	63.54	–	83.78	85.62	85.77^∗
NCBI-disease	1.77	57.18	57.18	–	68.00	86.89	–	86.89	87.82	88.25^∗
BC2GM	51.98	6.73	51.98	–	–	–	–	–	84.52	84.72^∗
JNLPBA	77.31	10.28	77.31	77.32	–	–	–	77.32	79.10	79.44^∗
Mean Score	27.20	48.14	70.60	77.32	81.08	75.22	89.84	84.86	86.08	86.28^∗

^∗

Highest performance for task (row).