Improve the Accuracy of Non-native Speech Annotation with a Semi-automatic Approach
Wei Wang, Wei Wei, Yanlu Xie, Minghao Guo, Jinsong Zhang · 2018
High accuracy feedback of Computer Aided Pronunciation Training system relies heavily on the availability of a substantial amount of high quality annotated non-native speech data. It is generally assumed that non-native annotation is very time and labor consuming with poorer quality. To address these problems, we compared two different transcription approaches in terms of the accuracy with which can describe the actual learner's speech better. The results showed that with similar inter-annotator agreements, the performance of semi-automatic approach was much better than manual approach.