MergeDistill: Merging Language Models using Pre-trained Distillation.
Simran KhanujaMelvin JohnsonPartha P. TalukdarPublished in: ACL/IJCNLP (Findings) (2021)
Keyphrases
- language model
- pre trained
- language modeling
- training data
- document retrieval
- n gram
- training examples
- probabilistic model
- information retrieval
- speech recognition
- language modelling
- retrieval model
- query expansion
- test collection
- relevance model
- control signals
- smoothing methods
- multimedia
- document ranking
- pairwise
- generative model
- support vector
- feature vectors
- information retrieval systems
- data sets
- statistical language models
- language models for information retrieval