Table 6.
Performance comparison in term of overall accuracy (%) of various Glove embedding dimensions on all datasets.
| Model | IMDB | MR (2002) | MR (2004) | (2005) |
|---|---|---|---|---|
| RAE [45] | - | - | - | 77.7 |
| MNB [46] | 86.6 | - | - | 79.0 |
| SVM [47] | 89.2 | - | - | - |
| NBSVM [46] | 91.2 | - | - | 79.4 |
| MTL-ML4 [51] | ||||
| LSTM [40] | 87.0 | 80.1 | ||
| BiLSTM [40] | 87.9 | - | - | 80.3 |
| AC-BiLSTM [40] | 91.8 | - | - | 83.2 |
| Static CNN [48] | - | - | - | 81.0 |
| Non-Static CNN [48] | - | - | - | 81.5 |
| Multichannel CNN [48] | 88.1 | 88.4 | 81.7 | 81.1 |
| BERT [49] | 94.0 | 88.0 | 89.0 | - |
| CNN [49] | 91.1 | 84.2 | 89.1 | - |
| B-CNN [49] | 93.0 | 87.1 | 90.3 | - |
| B-MLCNN [49] | 95.1 | 88.2 | 95.0 | - |
| Glove CNN-BiLSTM [38] | 89.5 | - | - | 84.4 |
| LSTM [50] | 81.7 | - | - | 72.7 |
| BiLSTM [50] | 85.0 | - | - | 74.7 |
| PEW (Proposed) | 86.4 | 81.2 | 83.4 | 85.9 |
| PEW + C (Proposed) | 86.9 | 83.1 | 84.1 | 84.1 |
| PEW + MC (Proposed) | 87.7 | 83.7 | 84.9 | 86.2 |
| PEW +MCAB (Proposed) | 90.3 | 84.1 | 85.9 | 87.1 |