BMC Medical Informatics and Decision Making

Table 2 Performance of systems on the clinical STS corpus of the BioCreative/OHNLP shared task in 2018

From: Distributed representation and one-hot representation fusion with gated network for clinical semantic textual similarity

Method	Score Interval
Method	[0,1]	[1, 2]	[2, 3]	[3, 4]	[4, 5]	Overall
Baseline
One-hot	0.5567	0.2311	0.0998	0.2409	0.1167	0.7939
CNN	0.3960	−0.0850	−0.0090	0.0370	−0.0654	0.4444
LSTM	0.3920	−0.2945	0.2088	−0.0538	−0.0303	0.4275
BERT	0.7613	0.1206	0.2635	0.2530	0.1210	0.8461
Concatenation
CNN + one-hot	0.5406	0.6917	0.1352	0.2539	0.0744	0.8083
LSTM+one-hot	0.5850	0.3415	0.2269	0.2173	0.2155	0.8030
BERT+one-hot	0.6684	0.3038	0.2309	0.2425	0.2203	0.8525
Fusion (gated network)
CNN + one-hot	0.6973	0.2324	0.1675	0.2336	0.0864	0.8442
LSTM+one-hot	0.6253	0.3583	0.1869	0.2550	0.1018	0.8379
BERT+one-hot	0.6872	0.1605	0.3238	0.2822	0.1666	0.8541

Back to article page

ISSN: 1472-6947

Contact us

General enquiries: journalsubmissions@springernature.com