Investigating Pitch Accent Recognition in Non-native Speech

Acquisition of prosody, in addition to vocabulary and grammar, is essential for language learners. However, it has received less attention in instruction. To enable automatic identification and feedback on learners' prosodic errors, we investigate automatic pitch accent labeling for nonnative speech. We demonstrate that an acoustic-based context model can achieve accuracies over 79% on binary pitch accent recognition when trained on withingroup data. Furthermore, we demonstrate that good accuracies are achieved in crossgroup training, where native and nearnative training data result in no significant loss of accuracy on non-native test speech. These findings illustrate the potential for automatic feedback in computer-assisted prosody learning.

• Type: Journal
• Year: 2009
• Where: ACL
• Authors: Gina-Anne Levow
• Keywords: ACL 2009, Acoustic-based Context Model, Computational Linguistics, Nearnative Training Data, Non-native Test Speech