Deduplicating Training Data Makes Language Models Better.
Katherine LeeDaphne IppolitoAndrew NystromChiyuan ZhangDouglas EckChris Callison-BurchNicholas CarliniPublished in: ACL (1) (2022)
Keyphrases
- language model
- training data
- language modeling
- document retrieval
- probabilistic model
- n gram
- retrieval model
- speech recognition
- statistical language models
- query expansion
- language modelling
- test collection
- information retrieval
- document ranking
- context sensitive
- pseudo relevance feedback
- smoothing methods
- learning algorithm
- decision trees
- query terms
- supervised learning
- training set
- ad hoc information retrieval
- labeled data
- vector space model
- word error rate
- language models for information retrieval
- translation model
- classification accuracy
- feature selection