Skip to main content

Table 4 Comparison between different RE hidden state augmentations

From: JCBIE: a joint continual learning neural network for biomedical information extraction

Corpus

Vanilla

Entity marker

Entity type embedding

Entity type prototype

\({\text {ADE}}_{1}\)

74.61

74.18

72.78

74.69

ADE

79.37

80.56

80.25

79.66

DDI + ADE

79.58

80.09

79.88

78.57

ADE + DDI + CPR

71.83

72.97

71.51

71.43

Avg.

76.35

76.95

76.10

76.09

  1. The bold means the best results
  2. The results are measured by RE micro-F1