MSR-86K: An Evolving, Multilingual Corpus with 86,300 Hours of Transcribed Audio for Speech Recognition Research.
Song LiYongbin YouXuezhi WangZhengkun TianKe DingGuanglu WanPublished in: CoRR (2024)
Keyphrases
- speech recognition
- spontaneous speech
- automatic speech recognition
- speaker identification
- broadcast news
- speech processing
- speech recognition technology
- speech signal
- conversational speech
- hidden markov models
- language model
- noisy environments
- speech recognizer
- speech synthesis
- acoustic features
- pattern recognition
- cepstral coefficients
- speech recognizers
- parallel corpus
- spoken language
- spoken document retrieval
- multimedia
- human machine interaction
- word recognition
- cross lingual
- speech retrieval
- audio visual speech recognition
- cross language
- signal processing
- speaker recognition
- language independent
- visual speech
- speaker dependent
- neural network
- parallel corpora
- statistical machine translation
- non stationary
- speaker adaptation
- feature vectors
- feature extraction
- image processing