Unsupervised domain discovery using latent dirichlet allocation for acoustic modelling in speech recognition.
Mortaza DoulatyOscar SazThomas HainPublished in: INTERSPEECH (2015)
Keyphrases
- speech recognition
- latent dirichlet allocation
- topic modeling
- speech recognition systems
- topic models
- speaker independent
- topic discovery
- lda model
- speech recognizers
- probabilistic topic models
- language model
- generative model
- acoustic models
- automatic speech recognition
- speech recognizer
- pattern recognition
- speech signal
- noisy environments
- hidden markov models
- hierarchical bayesian models
- text mining
- unsupervised learning
- speech synthesis
- knowledge discovery
- acoustic features
- neural network
- speaker identification
- prior knowledge
- mel frequency cepstral coefficients
- unsupervised manner
- feature extraction
- computer vision
- text analysis
- text documents
- text classification