Data-to-text generation on WebNLG Unseen v1
5.63Fluency ScoreOriginal
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| OriginalStatus=Human Reference2019.08 | 5.63 | — | — | 5.63 | |
| UPF-FORGE2019.08 | 5.51 | 35.7 | 0.37 | 5.24 | |
| GRUApproach=Pipeline2019.08 | 4.91 | 38.55 | 0.22 | 4.67 | |
| Melbourne2019.08 | 4.65 | 33.27 | 0.33 | 4.15 | |
| Majority2019.08 | 4.49 | 41.13 | 0.22 | 4.25 | |
| E2E TransformerApproach=End-to-End2019.08 | 4.32 | 5.88 | 0.09 | 3.81 | |
| Random2019.08 | 4.07 | 41.51 | — | 3.86 | |
| TransformerApproach=Pipeline2019.08 | 4.05 | 38.92 | 0.21 | 3.64 | |
| E2E GRUApproach=End-to-End2019.08 | 3.45 | 6.25 | 0.09 | 3.03 |