Prosody Training Mobile Application: Early Design Assessment and Lessons Learned

Computer-assisted prosody teaching is a relatively new topic of interest for software developers. Automatic speech recognition for the purposes of pronunciation instruction is an important technology within the context of technology-driven foreign language learning. This paper discusses an approach to extend audio perception by using visual perception channels represented as pitch graphs demonstrated to the learner. We discuss the pitch processing stage and the data processing algorithms required to construct smooth user-friendly pitch curves, as well as the pitch quality measures that can be used for formalized pitch quality evaluation. We introduce the iOS application as one of the end-user mobile components of a prosody-based learning framework and analyze the results of the early design proof-of-concept assessment stage using native speakers. Specifically, we analyze, why representing the conversational context is important to improve the prosody exercises and what is rationale to introduce the exercises able to explicitly differentiate between the different kinds of voice melody in the grammatical, emotional and attitudinal functions of intonation.