Automatic Pronunciation Error Detection and Feedback Generation for CALL Applications

Renlong Ai

In: Proceedings of the 17th International Conference on Human-Computer Interaction. Human Computer Interaction International Conferences (HCII-2015), August 2-7, Los Angeles, California, USA, Springer, 2015.


This paper describes a new method of automatic error detection in Computer Assisted Language Learning (CAPT) system. The method combines linguistic knowledge and modern speech technology. Our HMM classifier trained from annotations of linguists is not only capable of classifying correct and wrong phonemes, but also can tell how wrong an error phoneme is pronounced. Phone errors in L2’s speech, like phoneme substitution or distortion are detected with high accuracy, and at the same time, corrective feedback with multimedia support, which demonstrates how exactly error phonemes should be pronounced, is also generated.


error_detection.pdf (pdf, 627 KB )

German Research Center for Artificial Intelligence
Deutsches Forschungszentrum für Künstliche Intelligenz