Semantic-conditioned sentence generation on FrameNet After Filtering 1.7 (test)
97PPLHuman (FN 1.7)
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| Human (FN 1.7)2024.06 | 97 | 1 | 975 | |
| GPT-4 | FEConditioning=FE-Conditioning2024.06 | 103.4 | 0.826 | 838 | |
| GPT-4 | Frame + FEConditioning=Frame+FE-Conditioning2024.06 | 111.8 | 0.821 | 845 | |
| T5 | FEConditioning=FE-Conditioning2024.06 | 112.7 | 0.777 | 850 | |
| GPT-4Conditioning=None2024.06 | 114.2 | 0.723 | 724 | |
| T5Conditioning=None2024.06 | 117.5 | 0.713 | 789 | |
| T5 | Frame + FEConditioning=Frame+FE-Conditioning2024.06 | 124.4 | 0.704 | 873 | |
| Pancholy et al.2024.06 | 146 | 0.686 | 947 |