Unsupervised phoneme and word acquisition from continuous speech based on a hierarchical probabilistic generative model.
Masatoshi NaganoTomoaki NakamuraPublished in: Adv. Robotics (2023)
Keyphrases
- generative model
- semi supervised
- probabilistic model
- pitman yor process
- deep belief networks
- bayesian framework
- unsupervised learning
- mixture model
- topic models
- prior knowledge
- posterior probability
- conditional random fields
- maximum entropy principle
- hierarchical bayesian model
- em algorithm
- topic modeling
- discriminative learning
- latent dirichlet allocation
- markov chain monte carlo
- latent topics
- co occurrence
- speech recognition
- semi supervised learning
- n gram
- dirichlet process mixture models
- discriminative models
- hidden markov models
- learned models
- expectation maximization