Aspect-Based Sentiment Analysis on Restaurant-ACOS
86.12ATE (Accuracy)UNIFIEDABSA
Evaluation results
| Method | Links | ||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| UNIFIEDABSAModel Configuration=Multi-task instruction tuning (Unify), Backbone=T5-BASE, Evaluation Protocol=Fully-supervised2022.11 | 86.12 | 84.62 | 91.95 | 91.17 | 88.24 | 78.31 | 77.76 | 77.91 | 73.06 | 70.16 | 60.6 | 79.99 | |
| Single Dedicated ModelsModel Configuration=Single-task learning, Backbone=T5-BASE, Evaluation Protocol=Fully-supervised2022.11 | 84 | 83.04 | 90.66 | 89.69 | 86.81 | 76.33 | 73.27 | 75.81 | 68.81 | 68.22 | 59.38 | 77.82 | |
| UNIFIEDABSAshot=64-shot2022.11 | 70.43 | 70.09 | 87.41 | 84.77 | 79.22 | 61.77 | 56.87 | 61.64 | 53.71 | 47.74 | 36.75 | 64.58 | |
| Single Modelshot=64-shot2022.11 | 68.42 | 63.87 | 86.98 | 83.78 | 75.72 | 60.31 | 47.31 | 57.63 | 46.81 | 39.59 | 30.72 | 60.1 | |
| UNIFIEDABSAshot=32-shot2022.11 | 65.36 | 60.95 | 87.58 | 84.95 | 68.53 | 55.35 | 44.86 | 51.9 | 44.96 | 39 | 26.73 | 57.29 | |
| Single Modelshot=32-shot2022.11 | 61.63 | 57.21 | 84.06 | 83.38 | 68.29 | 48.57 | 34.26 | 47.12 | 34.2 | 27.21 | 18.84 | 51.34 |