Using Noisy Word-Level Labels to Train a Phoneme Recognizer based on Neural Networks by Expectation Maximization.
Chen LiBo ZhangShan HuangZhenhuan LiuPublished in: ICCPR (2019)
Keyphrases
- word level
- neural network
- expectation maximization
- word recognition
- speech recognition
- noisy environments
- hidden markov models
- em algorithm
- automatic speech recognition
- viterbi algorithm
- language independent
- document analysis
- machine translation
- pattern recognition
- document images
- probabilistic model
- character recognition
- n gram
- maximum likelihood
- word segmentation
- language model
- handwriting recognition
- sentence level
- generative model
- training data
- k means
- machine vision
- speech signal
- semantic roles
- clustering algorithm