End-to-End Aspect-Based Sentiment Analysis on REST SemEval 2015 2016 (test)
74.72F1 ScoreBERT+SAN
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| BERT+SANBackbone=bert-base-uncased2019.10 | 74.72 | 72.92 | 76.72 | |
| BERT+TFMBackbone=bert-base-uncased2019.10 | 74.41 | 72.39 | 76.64 | |
| BERT+CRFBackbone=bert-base-uncased2019.10 | 74.06 | 71.88 | 76.48 | |
| BERT+GRUBackbone=bert-base-uncased2019.10 | 73.24 | 70.61 | 76.2 | |
| BERT+LinearBackbone=bert-base-uncased2019.10 | 73.22 | 71.42 | 75.25 | |
| Luo et al., 2019Model Category=Existing Models2019.10 | 72.78 | — | — | |
| Li et al., 2019aModel Category=Existing Models2019.10 | 69.8 | 68.64 | 71.01 | |
| Liu et al., 2018Model Category=LSTM-CRF2019.10 | 66.38 | 68.46 | 64.43 | |
| Lample et al., 2016Model Category=LSTM-CRF2019.10 | 66.2 | 66.1 | 66.3 | |
| Ma and Hovy, 2016Model Category=LSTM-CRF2019.10 | 64.29 | 61.56 | 67.26 |