TY - GEN
T1 - Unsupervised prosodic break detection in Mandarin speech
AU - Huang, Jui Ting
AU - Hasegawa-Johnson, Mark
AU - Shih, Chilin
PY - 2008
Y1 - 2008
N2 - We propose that, in Mandarin speech, an automatic prosodic break detector can be trained without any prosodically labeled training data. We use only lexical and acoustic cues to create a small labeled training set, then use semi-supervised learning to train a prosodic break detector. A generative mixture model is proposed as the learning algorithm that can learn with both labeled and unlabeled data. The experiments in both English and Mandarin corpus verify our algorithm.
AB - We propose that, in Mandarin speech, an automatic prosodic break detector can be trained without any prosodically labeled training data. We use only lexical and acoustic cues to create a small labeled training set, then use semi-supervised learning to train a prosodic break detector. A generative mixture model is proposed as the learning algorithm that can learn with both labeled and unlabeled data. The experiments in both English and Mandarin corpus verify our algorithm.
UR - http://www.scopus.com/inward/record.url?scp=84890476051&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=84890476051&partnerID=8YFLogxK
M3 - Conference contribution
AN - SCOPUS:84890476051
SN - 9780616220030
T3 - Proceedings of the 4th International Conference on Speech Prosody, SP 2008
SP - 165
EP - 168
BT - Proceedings of the 4th International Conference on Speech Prosody
PB - International Speech Communication Association
T2 - 4th International Conference on Speech Prosody 2008, SP 2008
Y2 - 6 May 2008 through 9 May 2008
ER -