Active learning and semi-supervised learning for speech recognition: A unified framework using the global entropy reduction maximization criterion.
Dong YuBalakrishnan VaradarajanLi DengAlex AceroPublished in: Comput. Speech Lang. (2010)
Keyphrases
- speech recognition
- semi supervised learning
- active learning
- semi supervised
- labeled data
- unlabeled data
- supervised learning
- language model
- hidden markov models
- machine learning
- speech synthesis
- semi supervised classification
- labeled examples
- speech signal
- pattern recognition
- speech recognizer
- text categorization
- co training
- random sampling
- automatic speech recognition
- transfer learning
- training set
- training data
- learning algorithm
- training examples
- label propagation
- unsupervised learning
- labeled and unlabeled data
- unlabeled instances
- speech recognition systems
- neural network
- text classification
- pairwise constraints
- feature selection