Table 5 Result of argument analysis using deep learning model with attention mechanism (Word Embedding from Scratch)
From: Argument annotation and analysis using deep learning with attention mechanism in Bahasa Indonesia
No | Model name | Accuracy (%) | Recall (%) | Precision (%) | F1 macro (%) | ROC-AUC |
|---|---|---|---|---|---|---|
1 | CNN | 81.44 ± 2.54 | 77.33 ± 4.03 | 80.41 ± 2.73 | 78.18 ± 3.59 | 88.81 ± 0.02 |
2 | CNN + Att | 72.59 ± 6.41 | 65.74 ± 6.39 | 71.35 ± 7.65 | 65.89 ± 7.57 | 79.63 ± 0.05 |
3 | LSTM | 66.76 ± 1.88 | 51.14 ± 2.66 | 45.36 ± 19.95 | 42.47 ± 4.89 | 61.32 ± 0.03 |
4 | LSTM + Att | 72.71 ± 5.50 | 61.58 ± 9.63 | 60.05 ± 22.23 | 58.29 ± 15.21 | 79.03 ± 0.09 |
5 | GRU | 65.89 ± 1.87 | 54.21 ± 3.31 | 54.83 ± 11.50 | 51.35 ± 6.45 | 59.38 ± 0.04 |
6 | GRU + Att | 77.17 ± 6.30 | 71.37 ± 12.01 | 79.84 ± 3.43 | 69.34 ± 14.66 | 83.36 ± 0.10 |
7 | BiLSTM | 70.06 ± 5.66 | 62.56 ± 7.69 | 67.75 ± 6.46 | 61.14 ± 9.82 | 72.45 ± 0.06 |
8 | BiLSTM + Att | 75.32 ± 4.46 | 68.23 ± 8.05 | 74.10 ± 5.11 | 68.14 ± 8.94 | 79.65 ± 0.04 |
9 | BiGRU | 65.31 ± 2.56 | 57.45 ± 6.27 | 59.27 ± 6.03 | 55.61 ± 6.88 | 64.29 ± 0.05 |
10 | BiGRU + Att | 75.81 ± 5.78 | 69.52 ± 10.63 | 79.07 ± 3.94 | 67.55 ± 13.95 | 82.14 ± 0.10 |