Table 12 Detailed results achieved by the proposed MSTLTR Model and the Basic Model on DataMLEE. The Basic Model is trained only on the training and development sets of DataMLEE without transfer learning. MSTLTR Model is jointly trained on all available source domain datasets and the training and development sets of the target dataset DataMLEE using proposed transfer learning approach. The two models are tested on the test set of DataMLEE
From: A transfer learning model with multi-source domains for biomedical event trigger extraction
Trigger type | Basic model | Â | Â | MSTLTR model | Â | Â |
|---|---|---|---|---|---|---|
| Â | P | R | F1 | P | R | F1 |
Cell proliferation | 83.33 | 81.40 | 82.35 | 85.37 | 81.40 | 83.33 |
Development | 74.51 | 77.55 | 76.00 | 79.59 | 79.59 | 79.59 |
Blood vessel develop | 98.64 | 93.87 | 96.20 | 99.66 | 93.87 | 96.68 |
Growth | 88.89 | 85.71 | 87.27 | 91.23 | 92.86 | 92.04 |
Death | 66.67 | 81.08 | 73.17 | 74.36 | 78.38 | 76.32 |
Breakdown | 73.68 | 63.64 | 68.29 | 83.33 | 68.18 | 75.00 |
Remodeling | 75.00 | 30.00 | 42.86 | 83.33 | 50.00 | 62.50 |
Synthesis | 33.33 | 25.00 | 28.57 | 80.00 | 100.00 | 88.89 |
Gene expression | 85.40 | 88.64 | 86.99 | 89.13 | 93.18 | 91.11 |
Transcription | 50.00 | 16.67 | 25.00 | 100.00 | 50.00 | 66.67 |
Protein Catabolism | 0.0 | 0.0 | 0.0 | 100.00 | 20.00 | 33.33 |
Phosphorylation | 75.00 | 100.00 | 85.71 | 100.00 | 100.00 | 100.00 |
Dephosphorylation | 0.0 | 0.0 | 0.0 | 0.0 | 0.0 | 0.0 |
Localization | 76.81 | 79.70 | 78.23 | 86.61 | 82.71 | 84.62 |
Binding | 82.69 | 75.44 | 78.90 | 83.64 | 80.70 | 82.14 |
Regulation | 65.13 | 61.35 | 63.18 | 66.84 | 63.29 | 65.01 |
Positive regulation | 80.56 | 82.86 | 81.69 | 82.91 | 83.17 | 83.04 |
Negative regulation | 76.67 | 75.10 | 75.88 | 83.91 | 78.78 | 81.26 |
Planned process | 66.67 | 57.14 | 61.54 | 73.78 | 61.73 | 67.22 |
TOTAL | 79.47 | 77.23 | 78.34 | 83.96 | 79.89 | 81.88 |