A CALL system for oral proficiency is being developed in which constrained responses are elicited from L2 learners. In the first phase the best matching utterance is selected from a predefined list of possible responses. Since errors may occur and giving feedback on the basis of incorrectly recognized utterances is confusing, we verify the correctness of the utterance in the second phase. In the current paper we focus on the utterance verification process. Combining duration related features with a likelihood ratio (LR) yielded an equal error rate (EER) of 10.3%, which was significantly better than the EER for LR alone, 14.4%, and the EER for the duration-related features, 25.3%

Index Terms: utterance verification, non-native speech processing, computer-assisted language learning

@inproceedings{doremalen09_slate,
  title     = {Utterance verification in language learning applications},
  author    = {Joost van Doremalen and Helmer Strik and Catia Cucchiarini},
  year      = {2009},
  booktitle = {Speech and Language Technology in Education (SLaTE 2009)},
  pages     = {13--16},
  doi       = {10.21437/SLaTE.2009-4},
  issn      = {2311-4975},
}