Skip to main content

View full-text article in PMC

. Author manuscript; available in PMC: 2021 Aug 3.

Published in final edited form as: Proc ACM Conf Health Inference Learn (2020). 2020 Apr 2;2020:214–221. doi: 10.1145/3368555.3384455

Table 3:

Performance comparison of BERT_base against the model of Dernoncourt et al. for individual entities within the i2b2 2014 test corpus.

		Precision	Recall	F1
Entity type	Model
AGE	BERT_base	97.12	98.23	97.67
n = 789	Dernoncourt et al.	98.97	97.60	98.28
CONTACT	BERT_base	98.31	98.46	98.38
n = 648	Dernoncourt et al.	98.80	98.33	98.57
DATE	BERT_base	99.43	99.26	99.35
n = 8022	Dernoncourt et al.	99.06	99.52	99.29
ID	BERT_base	96.73	97.66	97.20
n = 1455	Dernoncourt et al.	99.29	98.76	99.02
LOCATION	BERT_base	97.14	94.12	95.60
n = 3027	Dernoncourt et al.	95.96	95.74	95.85
NAME	BERT_base	99.12	98.29	98.70
n = 5387	Dernoncourt et al.	98.22	99.15	98.68
PROFESSION	BERT_base	96.39	92.49	94.40
n = 346	Dernoncourt et al.	87.99	79.71	83.64