Abstract
Automatic prosody labelling is a useful front-end for automatic speech recognition, for automatic speech understanding, and for the development of corpora used to create speech synthesizers. Automatic labelling of prosody has also proven to be quite useful in the linguistic analysis of new speaking styles in a known language. This chapter provides a survey of the state-of-the-art best practices and open questions in the automatic labelling of prosodic information and its assessment. It describes the major prosodic inventories that are used in prosody labelling. It then discusses the relevance of acoustics and syntax in automatic labelling. A brief description of AuToBI, a tool that performs automatic ToBI labelling of US English, is provided. The chapter concludes by discussing methods of evaluating automatic prosody labelling.
Original language | English (US) |
---|---|
Title of host publication | The Oxford Handbook of Language Prosody |
Editors | Carlos Gussenhoven, Aoju Chen |
Publisher | Oxford University Press |
Pages | 646-656 |
Number of pages | 11 |
ISBN (Electronic) | 9780198832232 |
ISBN (Print) | 9780198832232 |
DOIs | |
State | Published - Dec 2020 |
Keywords
- acquisition
- word prosody
- lexical tone
- pitch accent
- word stress
- perpetual tuning
- word recognition
- word production
- developmental triggers
- prosodic information
- linguistic analysis
- automatic speech recognition
- acoustics
- automatic prosody labelling
- syntax
ASJC Scopus subject areas
- General Arts and Humanities
- General Social Sciences