Model-Generated Pretraining Signals Improves Zero-Shot Generalization of Text-to-Text Transformers.
Linyuan GongChenyan XiongXiaodong LiuPayal BajajYiqing XieAlvin CheungJianfeng GaoXia SongPublished in: ACL (1) (2023)
Keyphrases
- text retrieval
- formal model
- experimental data
- prior knowledge
- computational model
- information retrieval
- statistical model
- management system
- neural network
- free text
- parameter estimation
- text mining
- probabilistic model
- high level
- theoretical analysis
- semantic information
- similarity measure
- learning algorithm
- database
- key concepts
- analytical model
- statistical natural language processing