A novel language-adaptable accent reduction software template

摘要

This paper presents key elements of an Android mobile phone application designed to improve English speech fluency by accent reduction, with techniques designed to be generalizable to other languages and accessible to researchers with limited financial resources for implementation. After development, the software was implemented on a Samsung Galaxy S6, a common device for personal and business use. This paper will present existing approaches to this problem and associated issues, selection and use of existing open-source speech-recognition tools, and a general software architecture for using these tools in accent reduction applications. In particular, use of the CMU Sphinx software and associated corpi on the Android platform, changes that must be made to use the software in an accent reduction application, and practical considerations when designing an associated Graphical User Interface will be discussed. Finally, the Android Python port ”Kivy” and its practicality for this purpose will also be discussed. It was found that CMU Sphinx can run well on Android if implemented carefully, and be used to provide a rich user experience with both aural and visual feedback, although finding appropriate corpi for some languages can pose difficulty. Unfortunately, Kivy was found to be unreliable due to packaging errors.