Spoken CALL Shared Task system description

Andrew Caines


We describe the systems we applied to the Spoken CALL Shared Task text-processing track in which an 'accept' or 'reject' label had to be applied to transcribed responses to a given 'prompt' stimulus. The data come from a language tutoring system for Swiss-German learners of English. In our system we attempt to capture the grammaticality and semantic dimensions of language assessment: dimensions which are explicitly labelled in the training corpus. With the training data we achieved accuracies and differential scores well above the baseline. However, our system performed less well on the test data and we discuss ways to improve performance.


 DOI: 10.21437/SLaTE.2017-14

Cite as: Caines, A. (2017) Spoken CALL Shared Task system description. Proc. 7th ISCA Workshop on Speech and Language Technology in Education, 79-84, DOI: 10.21437/SLaTE.2017-14.


@inproceedings{Caines2017,
  author={Andrew Caines},
  title={ Spoken CALL Shared Task system description},
  year=2017,
  booktitle={Proc. 7th ISCA Workshop on Speech and Language Technology in Education},
  pages={79--84},
  doi={10.21437/SLaTE.2017-14},
  url={http://dx.doi.org/10.21437/SLaTE.2017-14}
}