Table 9.
Comparing the recalls of the long-tailed labels of TextCNN, TextRNN and TextRCNN with their CBS-improved models for automatic ICD coding.
| Model | Precision | Recall |
|---|---|---|
| TextCNN | 0.0188 | 0.0498 |
| TextRNN | 0.0169 | 0.0442 |
| TextRCNN | 0.0103 | 0.0269 |
| TextCNN+CBS | 0.2599 | 0.6369 |
| TextRNN+CBS | 0.2593 | 0.6617 |
| TextRCNN+CBS | 0.1943 | 0.4778 |