Keep Decoding Parallel With Effective Knowledge Distillation From Language Models To End-To-End Speech Recognisers.
Michael HentschelYuta NishikawaTatsuya KomatsuYusuke FujitaPublished in: ICASSP (2024)
Keyphrases
- end to end
- language model
- speech recognition
- language modeling
- n gram
- word error rate
- retrieval model
- probabilistic model
- test collection
- language modelling
- document retrieval
- statistical language models
- query expansion
- admission control
- information retrieval
- congestion control
- spoken term detection
- automatic speech recognition
- application layer
- speech signal
- vector space model
- context sensitive
- query terms
- scalable video
- transport layer
- language modeling framework
- language models for information retrieval
- relevance feedback