Table 2 Precision, recall and F1-score for both the corresponding test set and the respective other test set (i.e., cross evaluation)
From: We are not ready yet: limitations of state-of-the-art disease named entity recognizers
Algorithm | Train set | Test set | Precision[%] | Recall[%] | F1-Score[%] |
|---|---|---|---|---|---|
BioBERT | NCBI | NCBI | 84.62 | 90.09 | 87.27 |
BC5CDR | 69.77 | 67.75 | 68.75 | ||
BC5CDR | NCBI | 73.63 | 63.19 | 68.01 | |
BC5CDR | 82.07 | 85.39 | 83.07 | ||
TaggerOne | NCBI | NCBI | 83.46 | 82.66 | 83.06 |
BC5CDR | 70.01 | 40.75 | 51.51 | ||
BC5CDR | NCBI | 68.30 | 56.38 | 61.77 | |
BC5CDR | 83.59 | 80.67 | 82.11 | ||
scispaCy | BC5CDR | NCBI | 65.65 | 57.49 | 61.30 |
BC5CDR | 76.20 | 75.22 | 75.71 | ||
DNorm | NCBI | NCBI | 80.80 | 81.90 | 81.35 |
BC5CDR | 65.73 | 50.29 | 56.98 | ||
Stanza | NCBI | NCBI | 86.65 | 88.54 | 87.58 |
BC5CDR | 70.24 | 57.78 | 63.40 | ||
BC5CDR | NCBI | 75.57 | 62.50 | 68.42 | |
BC5CDR | 82.85 | 84.95 | 83.88 | ||
HUNER | NCBI | NCBI | 83.82 | 86.35 | 85.07 |
BC5CDR | 70.20 | 64.92 | 67.46 | ||
BC5CDR | NCBI | 77.84 | 69.90 | 73.66 | |
BC5CDR | 83.07 | 83.52 | 83.30 |