Differential Privacy, Linguistic Fairness, and Training Data Influence: Impossibility and Possibility Theorems for Multilingual Language Models.
Phillip RustAnders SøgaardPublished in: ICML (2023)
Keyphrases
- language model
- differential privacy
- training data
- differentially private
- language modeling
- cross lingual
- document retrieval
- privacy preserving
- n gram
- probabilistic model
- speech recognition
- retrieval model
- query expansion
- training set
- cross language
- test collection
- privacy preservation
- query terms
- smoothing methods
- data sharing
- translation model
- data sets
- language models for information retrieval
- digital libraries
- information retrieval
- cross language information retrieval
- data privacy
- natural language
- personal information
- relevance model
- natural language processing
- information extraction
- active learning
- search logs