Improving Distinction between ASR Errors and Speech Disfluencies with Feature Space Interpolation.
Seongmin ParkDongchan ShinSangyoun PaikSubong ChoiAlena KazakovaJihwa LeePublished in: CoRR (2021)
Keyphrases
- automatic speech recognition
- feature space
- spontaneous speech
- speech recognition
- recognition errors
- speech signal
- word error rate
- broadcast news
- human machine interaction
- conversational speech
- speech recognition errors
- speech retrieval
- hidden markov models
- speech corpus
- data points
- dimensionality reduction
- speech sounds
- noisy environments
- high dimensional
- spoken language
- input space
- spoken words
- high dimensionality
- feature selection
- classification accuracy
- input data
- error analysis
- speech recognizers
- feature extraction
- training set
- feature vectors
- low dimensional
- kernel function
- mean shift
- dimension reduction
- spoken document retrieval
- language model
- image representation
- gaussian mixture model
- data sets
- information retrieval