Table 10 Detailed performance results achieved by the proposed MSTLTR and the other leading trigger recognition systems
From: A transfer learning model with multi-source domains for biomedical event trigger extraction
Trigger recognition system | Precision | Recall | F1-measure |
|---|---|---|---|
SVM feature-based System [8] | 81.44 | 69.48 | 75.67 |
SVM-based System [31] | 80.60 | 81.60 | 78.32 |
Neural Network-based System [32] | 71.04 | 84.60 | 77.23 |
CNN-based Neural Network System [33] | 80.67 | 76.76 | 78.67 |
RNN-based Neural Network System [34] | 79.78 | 78.45 | 79.11 |
Attention-based Neural Network System [35] | 81.33 | 79.48 | 80.39 |
Tree-base Neural Network System [11] | 81.12 | 79.15 | 80.28 |
Convolutional Highway Neural Network System [12] | 80.06 | 81.25 | 80.57 |
Hybrid Neural Network System [13] | 80.03 | 81.54 | 80.66 |
Joint-GATE-Sentence Neural Network System [36] | 81.58 | 81.08 | 81.33 |
Joint-GATE-Document Neural Network System [36] | 82.11 | 82.53 | 82.32 |
BioBERT-based Neural Network System [37] | 79.48 | 83.76 | 81.57 |
Our MSTLTR System | 83.96 | 79.89 | 81.88 |