Abstract
In this paper, we propose a machine scoring method for imitation reading in Computer-based English Listening and Speaking Test (CELST). It focuses on how to evaluate the quality of paragraph reading. First of all, the paragraph is divided into sentences. After that, the individual sentences are scored from the dimensions of pronunciation accuracy, accent, rhythm, intonation and speech rate. Third, the score value of each sub-indicator is obtained by scoring the average value of the pronunciation accuracy of the plurality of sentences in the paragraph, the mean value of the accent score, the mean value of the rhythm score, the mean value of the intonation score and the mean value of the speech rate score. Finally, we determined the weight of each sub-indicator by consulting experts to obtain the assessment model. Experiment shows that our system scoring and teacher scoring are in good agreement. If the system is applied to help candidates to self-train reading imitation, it is expected to improve the candidates ability to imitate the paragraph reading.
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have