Training Hybrid Models on Noisy Transliterated Transcripts for Code-Switched Speech Recognition.
Matthew WiesnerMousmita SarmaAshish AroraDesh RajDongji GaoRuizhe HuangSupreet PreetMoris JohnsonZikra IqbalNagendra GoelJan TrmalLeibny Paola García-PereraSanjeev KhudanpurPublished in: Interspeech (2021)
Keyphrases
- speech recognition
- hybrid models
- automatic speech recognition
- noisy environments
- speech recognizers
- wall street journal corpus
- feed forward neural networks
- hidden markov models
- hybrid model
- isolated word
- acoustic models
- speech recognizer
- language model
- pattern recognition
- speech synthesis
- speech signal
- neural network
- discriminative training
- speech processing
- speech recognition technology
- speaker identification
- training set
- speech recognition systems
- multi layer
- broadcast news
- text classification
- speaker dependent
- speaker independent
- training phase
- audio visual speech recognition
- artificial intelligence