Abstract

This study develops an oral evaluation system of voice skills in English public speech passages and examines the system’s robustness. The system builds on four methods: 1) creation of customized oral practice system on the cloud; 2) automatic speech-text alignment; 3) pause detection; and 4) English speech scoring. The system was made available as a web application and was able to rate the vocal expressiveness and pronunciation correctness of the users’ public speech recordings. The application gave instant feedbacks of scores of pronunciation, duration, and intonation on the word level as well as a score indicating the overall performance of the speech passage. The system and a human judge rated sixty students’ oral public speech recordings respectively and produced two sets of scores. A paired t-test was executed on the two sets of scores to verify the robustness of the system. The result shows that the two ratings are not significantly different. It means that the automatic system can be an alternative to the manual scoring of a human judge in distance learning. In conclusion, the system can serve as an additional evaluative aid for teachers and self-evaluation aid for English public speech learners. The success of the application implies that it may be possible for researchers to further develop the system for different language settings and ultimately becomes a multilingual evaluative tool.