Skip to main content

Table 3 Micro-averaged performance for chemical and drug recognition task using BiLSTM-CRF approach

From: Combining word embeddings to extract chemical and drug entities in biomedical literature

 

Precision (%)

Recall (%)

F1-score (%)

Based on BERT (Xiong et al. [61])

91.23

90.88

91.05

Classic WE + Contextual WE + Medical WE

91.41

90.14

90.77

Medical WE

87.94

86.24

87.08

Contextual WE

88.74

85.22

86.95

Classic WE

86.53

83.46

84.96

CRF + features (López-Úbeda et al. [6])

88.51

69.81

78.06