Which resemblance is useful to predict phrase boundary rise labels for Japanese expressive text-to-speech synthesis, numerically-expressed stylistic or distribution-based semantic?
Hideharu NakajimaHideyuki MizunoOsamu YoshiokaSatoshi TakahashiPublished in: INTERSPEECH (2013)
Keyphrases
- text to speech synthesis
- text to speech
- semantic classes
- pairwise
- semantic knowledge
- natural language
- semantic information
- training data
- domain specific
- training examples
- multi label
- domain ontology
- random variables
- class membership
- semantic web technologies
- spatial distribution
- japanese language
- syntactic analysis
- semantic annotation
- power law
- semantic features
- semantic network
- search engine
- class labels
- semantic web
- high level