P3LM: Probabilistically Permuted Prophet Language Modeling for Generative Pre-Training.
Junwei BaoYifan WangYing JiangyongYeyun GongJing ZhaoYouzheng WuXiaodong HePublished in: EMNLP (Findings) (2022)
Keyphrases
- language modeling
- language model
- retrieval model
- information retrieval
- cross lingual
- query expansion
- probabilistic model
- discriminative models
- n gram
- language modeling framework
- generative model
- text classification
- training set
- document retrieval
- retrieval effectiveness
- relevance model
- statistical language models
- data analysis
- pseudo relevance feedback
- digital libraries
- sentence retrieval
- improvements in retrieval effectiveness
- dirichlet prior
- unsupervised learning
- training data