Table 2.
Macro-averaged performance values in terms of precision, recall and F1-score for baseline (B) and transfer learning (TL) over the corpora per each entity type
| Precision (%) |
Recall (%) |
F1-score (%) |
||||
|---|---|---|---|---|---|---|
| B | TL | B | TL | B | TL | |
| Chemicals | 87.10 | 87.05 | 89.19 | 89.47 | 88.08 | 88.21 |
| Diseases | 80.41 | 81.41 | 81.13 | 82.46 | 80.73 | 82.09 |
| Species | 84.18 | 84.52 | 84.44 | 90.12 | 84.20 | 87.01 |
| Genes/proteins | 82.09 | 83.38 | 80.85 | 83.08 | 81.20 | 83.09 |
Note: Baseline values are derived from training on the target dataset only, while transfer learning values are derived by training on the source dataset followed by training on the target dataset. The macro average is computed by averaging the performance scores obtained by the classifiers for each corpus of a given entity class. Bold: best scores.