Language Modeling on One Billion Word corpus 1M sentences
71.8PerplexityURNNG
Evaluation results
| Method | Links | |
|---|---|---|
| URNNG2019.04 | 71.8 | |
| RNNG* -> URNNGTraining protocol=Supervised RNNG pre-training followed by URNNG fine-tuning2019.04 | 72 | |
| RNNG*Training source=Predicted parse trees from Kitaev and Klein (2018)2019.04 | 72.9 | |
| RNNLM2019.04 | 77.4 | |
| PRPNHyperparameters=Tuned by authors of this work2019.04 | 77.7 |