Feature selection for log-linear acoustic models.
Simon WieslerAlexander RichardYotaro KuboRalf SchlüterHermann NeyPublished in: ICASSP (2011)
Keyphrases
- log linear
- discriminative training
- acoustic models
- feature selection
- support vector machine
- hidden markov models
- log linear models
- maximum likelihood
- probabilistic interpretation
- text categorization
- posterior probability
- speech recognition
- efficient learning
- feature set
- multi class
- text classification
- automatic speech recognition
- machine learning
- probabilistic modeling
- co training
- feature extraction
- latent variables
- max margin
- support vector
- knn
- probabilistic model
- feature space
- model selection
- multi modal
- learning algorithm