LAE-ST-MoE: Boosted Language-Aware Encoder Using Speech Translation Auxiliary Task for E2E Code-switching ASR.
Guodong MaWenxuan WangYuke LiYuting YangBinbin DuHaoran FuPublished in: CoRR (2023)
Keyphrases
- automatic speech recognition
- speech recognition
- speech signal
- spoken language
- target language
- spontaneous speech
- text to speech
- language resources
- broadcast news
- machine translation
- language acquisition
- programming language
- english text
- error control
- word error rate
- text to speech synthesis
- bilingual dictionaries
- conversational speech
- source language
- machine translation system
- source code
- natural language
- speech corpus
- noisy environments
- spoken words
- language learning
- programs written
- bit rate
- speech sounds
- cross language information retrieval
- spoken dialog systems
- audio visual
- speech synthesis
- rate distortion
- low complexity
- error correction
- language processing
- comparable corpora
- human communication
- recognition errors
- language generation
- error detection
- cross language
- video compression