Table 4. F1-score results on each label of our proposed ConBGAT model with three compared baseline models on the SROIE dataset.
Entities | Spectral graph convolutional network | Bi-LSTM-CRF | BERT-CRF | ConBGAT |
---|---|---|---|---|
Company | 0.85 | 0.851 | 0.868 | 0.98 |
Address | 0.93 | 0.883 | 0.891 | 0.98 |
Date | 0.95 | 0.942 | 0.962 | 0.97 |
Total | 0.93 | 0.835 | 0.847 | 0.95 |
Macro average | 0.915 | 0.878 | 0.892 | 0.97 |