Generation from meaning representations on E2E (test)
0.686BLEUreconstructor-based pragmatic system (SR)
Evaluation results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| reconstructor-based pragmatic system (SR)approach=reconstructor-based pragmatic2019.04 | 0.686 | 0.0873 | 0.4525 | 0.7082 | 0.0237 | |
| distractor-based pragmatic system (SP)approach=distractor-based pragmatic2019.04 | 0.6776 | 0.0872 | 0.4459 | 0.6941 | 0.0227 | |
| S0type=base model2019.04 | 0.6652 | 0.0855 | 0.4445 | 0.6934 | 0.0223 | |
| Best Prev. (E2E Challenge)source=Ensemble of various challenge winners2019.04 | 0.6619 | 0.0861 | 0.4529 | 0.7083 | 0.0227 | |
| T-Genstatus=baseline2019.04 | 0.6593 | 0.0861 | 0.4483 | 0.685 | 0.0223 | |
| S0 x2type=ensemble, ensemble_size=2 models2019.04 | 0.6593 | 0.0831 | 0.4352 | 0.6958 | 0.0212 |