Table 7.
Correct classification rates (in %) of the ablation study.
| Experiment | Single component | Two components | Three components | |||||
|---|---|---|---|---|---|---|---|---|
| Polarity | Bi-LSTM | Logistic reg | (Polarity + Bi-LSTM) | (Polarity + logistic reg) | (Bi-LSTM + logistic reg) | Proposed framework | ||
| Experiment 1: Coronavirus-tagged data | 3-class | 87.15 | 85.57 | 87.67 | 87.94 | 88.20 | 86.15 | 89.31 |
| Experiment 2: Google Play reviews | 2-class | 87.44 | 91.33 | 90.83 | 88.40 | 88.43 | 90.61 | 92.87 |
| Experiment 3: Amazon Alexa | 2-class | 83.27 | 82.33 | 88.19 | 86.27 | 84.56 | 86.23 | 86.72 |
| Experiment 4: Rotten Tomatoes | 2-class | 76.28 | 81.47 | 81.53 | 80.41 | 81.77 | 82.15 | 83.48 |