Skip to main content

Table 5 Comparison of the performances of various systems on the three tasks (%)

From: Entity recognition from clinical texts via recurrent neural network

 

System

Method

Exact F1-score

Inexact F1-score

2010

LSTM + char-LSTM

RNN

85.81

92.91

Tang et al (2013) [10]

SSVM

85.82

92.40

Bruijin et al (2011)* [13]

Semi-Markov

85.23

92.44

Kim et al (2015) [9]

CRFs

84.30

-

Jiang et al (2011) [12]

CRFs

83.91

91.30

 

System

Method

Span F1-score

Type Accuracy

2012

LSTM + char-LSTM

RNN

92.29

86.94

Xu et al. (2013)* [15]

CRFs

91.66

85.74

Tang et al. (2013) [16]

CRFs + SVM

90.13

83.60

Sohn et al. (2013) [17]

CRFs

87.00

76.77

Aleksandar et al. (2013) [18]

CRFs

87.29

82.00

 

System

Method

Exact F1-score

Token F1-score

2014

LSTM + char-LSTM

RNN

94.29

96.54

Yang et al. (2015) [20]

CRFs

93.60

96.11

He et al. (2015) [22]

CRFs

92.32

95.14

Liu et al. (2015) [21]

CRFs + rule

91.24

94.64

Dehghan et al. (2015) [23]

CRFs + rule

91.13

95.31