Pitch tracking and tone features for Mandarin speech recognition

This is the first variationist sociotonetic study to use free-speech data for exploring tone. Due to the challenges of analyzing tone in free-speech data, prior work on sociotonetics has been limited to relatively formal speech styles: word lists, sentence frames, and phrase lists. But connected speech styles, including free speech and reading passages, are important for segmental sociophonetics and most other linguistic variables. Will free-speech data always be out of reach for sociotonetics? Can tone variation in connected speech data be normalized and meaningfully analyzed for sociolinguistic research questions? Using field data from the Sui language of China, this paper develops a practical approach for analyzing tone variation in connected speech data, and then applies it to a specific research question about dialect contact in exogamous Sui villages. Results show that some types of intra- and inter-speaker tone variation in connected speech can be effectively analyzed, although other types of tone variables are neutralized in this speech style.

show abstract

Pitch tracking and tone features for Mandarin speech recognition

Cited by 32 publications

References 14 publications

Tone Modeling for Continuous Mandarin Speech Recognition

Tone Modeling for Continuous Mandarin Speech Recognition

An Innovative Prosody Modeling Method for Chinese Speech Recognition

Sociotonetics using connected speech

Contact Info

Product

Resources

About