Table 3.
Performance of event classification for the test set.
Model | Overall (Micro) | Disposition (Strict) | No Disposition (Strict) | Undetermined (Strict) | ||||||
---|---|---|---|---|---|---|---|---|---|---|
Pre | Rec | F1 | Pre | Rec | F1 | Pre | Rec | F1 | ||
GatorTron | 0.9379 | 0.8782 | 0.8671 | 0.8726 | 0.9648 | 0.9655 | 0.9652 | 0.6911 | 0.7025 | 0.6967 |
GatorTronS | 0.9362 | 0.8490 | 0.8232 | 0.8359 | 0.8893 | 0.9310 | 0.9097 | 0.7258 | 0.5172 | 0.6040 |
RoBERTa | 0.8588 | 0.8111 | 0.7374 | 0.7725 | 0.8346 | 0.9255 | 0.8777 | 0.6600 | 0.3793 | 0.4818 |
RoBERTa MIMIC | 0.9251 | 0.8323 | 0.8797 | 0.8554 | 0.9646 | 0.9609 | 0.9628 | 0.7383 | 0.6529 | 0.6930 |
ALBERT | 0.8472 | 0.8111 | 0.7374 | 0.7725 | 0.8346 | 0.9255 | 0.8777 | 0.6600 | 0.3793 | 0.4818 |
ALBERT MIMIC | 0.9179 | 0.8012 | 0.8797 | 0.8386 | 0.9666 | 0.9533 | 0.9599 | 0.7103 | 0.6281 | 0.6667 |
Best precision, recall, and F1-scores are highlighted in bold.