Table 3 Comparison of classification accuracy and Macro-F1 for ALSC
From: Aspect-level sentiment classification with fused local and global context
Methods | Restaurant | Laptop | MAMS | |||||
|---|---|---|---|---|---|---|---|---|
Accuracy | Macro-F1 | Accuracy | Macro-F1 | Accuracy | Macro-F1 | Accuracy | Macro-F1 | |
TD-LSTM | 75.63 | – | 68.13 | – | 70.80 | 69.00 | – | – |
IAN | 78.60 | – | 72.10 | – | – | – | – | – |
MemNet | 78.16 | 65.83 | 70.33 | 64.09 | – | – | – | – |
RAM | 80.23 | 70.80 | 74.49 | 71.35 | 69.36 | 67.30 | – | – |
BERT | 82.86 | 74.87 | 77.12 | 72.55 | 74.42 | 72.67 | 81.96 | 81.28 |
BERT-SPC | 84.46 | 76.98 | 78.99 | 75.03 | 73.55 | 72.14 | 82.82 | 81.90 |
BERT-PT | 84.95 | 76.96 | 78.07 | 75.08 | – | – | – | – |
BAT | 86.03 | 79.24 | 79.35 | 76.50 | – | – | – | – |
T-GCN + BERT | 86.16 | 79.95 | 80.88 | 77.03 | 76.45 | 75.25 | 83.38 | 82.77 |
dotGCN + BERT | 86.16 | 80.49 | 81.03 | 78.10 | 78.11 | 77.00 | – | – |
PConvBERT (ours) | 86.96 | 80.87 | 81.66 | 78.33 | 76.73 | 75.82 | 84.36 | 83.95 |
RoBERTa | 87.23 | 80.20 | 81.19 | 77.69 | 74.58 | 72.75 | 84.06 | 83.45 |
PConvRoBERTa (ours) | 89.29 | 84.27 | 83.54 | 80.89 | 78.47 | 77.53 | 85.55 | 85.05 |