How Much Do Language Models Copy From Their Training Data? Evaluating Linguistic Novelty in Text Generation Using RAVEN.
R. Thomas McCoyPaul SmolenskyTal LinzenJianfeng GaoAsli CelikyilmazPublished in: Trans. Assoc. Comput. Linguistics (2023)
Keyphrases
- language model
- text generation
- training data
- natural language generation
- language modeling
- natural language
- n gram
- information retrieval
- document retrieval
- speech recognition
- probabilistic model
- test collection
- retrieval model
- query expansion
- language modelling
- language models for information retrieval
- training set
- learning algorithm
- statistical language models
- novelty detection
- ad hoc information retrieval
- relevance model
- theorem prover
- context sensitive
- pseudo relevance feedback
- translation model
- classification accuracy
- decision trees
- class labels
- retrieval effectiveness
- smoothing methods
- okapi bm
- naive bayes
- vector space model
- natural language processing
- document ranking
- artificial intelligence
- information retrieval systems
- labeled data