LICHEE: Improving Language Model Pre-training with Multi-grained Tokenization

34064

Knowledge Triples

Source Entity	Relationship	Target Entity
Fully Non-Autoregressive Neural Machine Translation Method	achieves	BLEU Metric
Fully Non-Autoregressive Neural Machine Translation Method	adopts	Dependency Reduction Method
Fully Non-Autoregressive Neural Machine Translation Method	comparesWith	Transformer Baseline Method