Table 6 Performance comparison (F1 score) of our method that uses the MKL as combination mechanism against other systems, including the recurrent architecture proposed by Crichton et al. [4], OGER, and fastText
From: Learning adaptive representations for entity recognition in the biomedical domain
Entity | Our sys. | Crichton et al. | OGER | fastText |
|---|---|---|---|---|
chebi | 78.99 | 74.83 | 53.73 | 75.24 |
cell | 80.16 | 86.89 | 79.91 | 80.03 |
go_cc | 68.92 | 63.08 | 54.63 | 67.99 |
go_bpmf | 36.20 | - | 26.81 | 30.27 |
organism | 94.99 | 97.44 | 44.11 | 92.64 |
protein | 84.38 | 75.16 | 62.00 | 80.76 |
sequence | 75.08 | - | 53.53 | 72.76 |