ComSL: A Composite Speech-Language Model for End-to-End Speech-to-Text Translation.
Chenyang LeYao QianLong ZhouShujie LiuYanmin QianMichael ZengXuedong HuangPublished in: NeurIPS (2023)
Keyphrases
- end to end
- language model
- speech recognition
- translation model
- cross language retrieval
- word error rate
- out of vocabulary
- language modeling
- statistical machine translation
- n gram
- document retrieval
- speech signal
- probabilistic model
- information retrieval
- automatic speech recognition
- query expansion
- language modelling
- context sensitive
- retrieval model
- query terms
- congestion control
- test collection
- mixture model
- statistical language models
- ad hoc information retrieval
- document length
- cross language information retrieval
- smoothing methods
- pseudo relevance feedback
- query translation
- question answering
- generative model
- co occurrence
- language model for information retrieval