Regularization, adaptation, and non-independent features improve hidden conditional random fields for phone classification.
Yun-Hsuan SungConstantinos BoulisChristopher D. ManningDan JurafskyPublished in: ASRU (2007)
Keyphrases
- classification accuracy
- feature set
- classification method
- feature vectors
- feature extraction
- feature space
- classification models
- classification process
- svm classifier
- extracted features
- support vector machine
- pattern recognition
- decision trees
- class labels
- support vector machine svm
- hidden conditional random fields
- text classification
- support vector
- multiscale
- image features
- feature subset
- texture analysis
- computer vision
- training set
- feature selection