Deduplicating Training Data Mitigates Privacy Risks in Language Models.
Nikhil KandpalEric WallaceColin RaffelPublished in: CoRR (2022)
Keyphrases
- language model
- training data
- language modeling
- n gram
- language modelling
- probabilistic model
- document retrieval
- learning algorithm
- training set
- retrieval model
- information retrieval
- speech recognition
- test collection
- supervised learning
- context sensitive
- decision trees
- statistical language models
- labeled data
- language model for information retrieval
- language models for information retrieval
- smoothing methods
- classification accuracy
- pseudo relevance feedback
- document ranking
- query terms
- ad hoc information retrieval
- vector space model
- passage retrieval
- translation model
- cross lingual
- query expansion
- naive bayes