PRINCE: Prefix-Masked Decoding for Knowledge Enhanced Sequence-to-Sequence Pre-Training

31313

Knowledge Triples

Source Entity	Relationship	Target Entity
Data to Text Generation Task	evaluatedBy	BLEU Metric
Denoising Autoencoder Method	enhances	Knowledge Enhanced Decoder Method
Knowledge Enhanced Decoder Method	appliesTo	Data to Text Generation Task
Knowledge Enhanced Decoder Method	predicts	Entity Token OtherEntity
Pre-trained Language Models Method	includes	Denoising Autoencoder Method