LAMASSU: A Streaming Language-Agnostic Multilingual Speech Recognition and Translation Model Using Neural Transducers.
Peidong WangEric SunJian XueYu WuLong ZhouYashesh GaurShujie LiuJinyu LiPublished in: INTERSPEECH (2023)
Keyphrases
- speech recognition
- translation model
- language model
- comparable corpora
- machine translation system
- bilingual dictionaries
- cross lingual
- language modeling
- cross language information retrieval
- parallel corpus
- chinese english
- isolated word
- statistical machine translation
- hidden markov models
- probabilistic model
- retrieval model
- document retrieval
- automatic speech recognition
- information retrieval
- query expansion
- n gram
- speech signal
- speech recognition systems
- pattern recognition
- context sensitive
- query translation
- machine translation
- statistical models
- digital libraries
- test collection
- neural network
- co occurrence
- parallel corpora
- speaker identification
- wordnet
- language independent
- news articles
- training data
- language processing
- maximum likelihood
- text classification
- statistical model
- query terms
- transfer learning