AdaMER-CTC: Connectionist Temporal Classification with Adaptive Maximum Entropy Regularization for Automatic Speech Recognition.
SooHwan EomEunseop YoonHee Suk YoonChanwoo KimMark Hasegawa-JohnsonChang D. YooPublished in: ICASSP (2024)
Keyphrases
- maximum entropy
- automatic speech recognition
- maximum entropy discrimination
- class conditional
- maximum entropy principle
- speech recognition
- pattern recognition
- text classification
- neural network
- iterative scaling
- principle of maximum entropy
- kernel logistic regression
- image classification
- conditional random fields
- machine learning
- bregman divergences
- hidden markov models
- model selection
- transformation based learning
- feature space
- speech signal
- support vector machine
- classification accuracy
- broadcast news
- feature vectors
- feature selection
- feature extraction
- support vector
- conversational speech
- multi task
- higher order
- em algorithm
- segmentation method
- energy function