On the Problem of Text-To-Speech Model Selection for Synthetic Data Generation in Automatic Speech Recognition.
Nick RossenbachRalf SchlüterSakriani SaktiPublished in: CoRR (2024)
Keyphrases
- model selection
- data generation
- automatic speech recognition
- text to speech
- speech synthesis
- speech recognition
- speech corpus
- cross validation
- speech signal
- active learning
- hyperparameters
- broadcast news
- hidden markov models
- data streams
- feature selection
- regression model
- word processing
- machine learning
- mixture model
- high throughput
- model selection criteria
- streaming data
- noisy environments
- selection criterion
- information criterion
- pattern recognition
- gaussian process
- co training
- computer vision
- data sets
- non stationary