Non-autoregressive Mandarin-English Code-switching Speech Recognition with Pinyin Mask-CTC and Word Embedding Regularization.
Shun-Po ChuangHeng-Jui ChangSung-Feng HuangHung-yi LeePublished in: CoRR (2021)
Keyphrases
- speech recognition
- autoregressive
- chinese characters
- speech recognizer
- speech recognition systems
- error tolerant
- text input
- speech recognition technology
- non stationary
- broadcast news
- speaker independent
- english text
- hidden markov models
- language model
- automatic speech recognition
- character recognition
- random fields
- speaker identification
- n gram
- speech signal
- speech synthesis
- language identification
- natural language
- spoken document retrieval
- noisy environments
- pattern recognition
- word level
- word recognition
- co occurrence
- sar images
- word sense disambiguation
- vector space
- information retrieval
- cross language information retrieval
- out of vocabulary
- machine vision
- maximum likelihood
- neural network