From: Toward multi-label sentiment analysis: a transfer learning based approach
Model | Accuracy (%) | Hamming loss | Macro F1 | Micro F1 |
---|---|---|---|---|
Proposed models | ||||
 BERT | 61.65 | 0.032 | 0.48 | 0.70 |
 XLNet | 66.61 | 0.027 | 0.56 | 0.77 |
Baseline deep learning models | ||||
 LSTM | 35.66 | 0.053 | 0.21 | 0.49 |
 BiLSTM | 36.88 | 0.051 | 0.25 | 0.49 |
 CNN + LSTM | 19.20 | 0.056 | 0.08 | 0.33 |
Baseline machine learning models | ||||
 SGD + OR | 28.69 | 0.052 | 0.26 | 0.47 |
 LR + OR | 15.35 | 0.051 | 0.12 | 0.29 |
 SVC + OR | 27.72 | 0.049 | 0.24 | 0.45 |
 RF + OR | 16.21 | 0.051 | 0.14 | 0.31 |
 SGD + BR | 28.98 | 0.051 | 0.26 | 0.47 |
 LR + BR | 15.35 | 0.051 | 0.12 | 0.29 |
 SVC + BR | 27.72 | 0.049 | 0.24 | 0.45 |
 RF + BR | 16.21 | 0.051 | 0.14 | 0.31 |
 SGD + CC | 39.29 | 0.055 | 0.29 | 0.49 |
 LR + CC | 31.67 | 0.056 | 0.16 | 0.43 |
 SVC + CC | 41.12 | 0.053 | 0.26 | 0.51 |
 RF + CC | 25.09 | 0.051 | 0.16 | 0.40 |
 SGD + LP | 38.09 | 0.062 | 0.29 | 0.45 |
 LR + LP | 39.00 | 0.060 | 0.20 | 0.45 |
 SVC + LP | 40.44 | 0.059 | 0.29 | 0.47 |
 RF + LP | 37.92 | 0.062 | 0.29 | 0.45 |
 SGD + RakelD | 38.09 | 0.062 | 0.29 | 0.45 |
 LR + RakelD | 39.00 | 0.060 | 0.20 | 0.45 |
 SVC + RakelD | 40.44 | 0.059 | 0.29 | 0.47 |
 RF + RakelD | 37.92 | 0.062 | 0.24 | 0.44 |
 BRkNNa | 5.27 | 0.103 | 0.01 | 0.09 |
 BRkNNb | 24.16 | 0.060 | 0.19 | 0.36 |
 MLARAM | 20.50 | 0.080 | 0.02 | 0.26 |
 MLkNN | 24.86 | 0.054 | 0.08 | 0.31 |