The Distributional Hypothesis Does Not Fully Explain the Benefits of Masked Language Model Pretraining.
Ting-Rui ChiangDani YogatamaPublished in: EMNLP (2023)
Keyphrases
- language model
- language modeling
- n gram
- information retrieval
- document retrieval
- speech recognition
- query expansion
- probabilistic model
- language modelling
- retrieval model
- statistical language models
- query terms
- smoothing methods
- context sensitive
- co occurrence
- test collection
- vector space model
- statistical machine translation
- mixture model
- language model for information retrieval
- word error rate
- ad hoc information retrieval
- pseudo relevance feedback
- generative model