An automatic feedback system for English speaking integrating pronunciation and prosody assessments

We have proposed a computer-assisted language learning (CALL) system, called Postech English Speaking Assessment and Assistant (PESAA) for non-native English learners, especially Koreans, to improve their overall language skills. PESAA is an automatic feedback system for speaking English that integrates both pronunciation and prosody assessments. The system has three error-feedback modules: the pronunciation, rhythm, and phrase break error-feedback modules. PESAA generates scores on each of three modules as well as a combined score. The pronunciation assessment gives feedback that is based on comparing canonical and actual phoneme alignment results. The rhythm and phrase break assessments give feedback that is based on comparing predictions with detected results. English learners can use PESAA to practice pronunciations, rhythm and phrase breaks by themselves. We evaluated PESAA in three different ways: accuracy, correlation of the system’s assessment with human assessments, and user satisfaction from the expected learning effectiveness and user interface. The evaluation showed that PESAA could work as a CALL system well, with good accuracy and positive learning results for the users.