RoAST: Robustifying Language Models via Adversarial Perturbation with Selective Training.
Jaehyung KimYuning MaoRui HouHanchao YuDavis LiangPascale FungQifan WangFuli FengLifu HuangMadian KhabsaPublished in: EMNLP (Findings) (2023)
Keyphrases
- language model
- language modeling
- probabilistic model
- document retrieval
- n gram
- language modelling
- information retrieval
- speech recognition
- context sensitive
- test collection
- statistical language models
- retrieval model
- smoothing methods
- ad hoc information retrieval
- query expansion
- language models for information retrieval
- training set
- document ranking
- okapi bm
- pseudo feedback
- query terms
- translation model
- relevance model
- vector space model
- text mining
- active learning
- training data