Edinburgh Napier University

  This project aims to address pronunciation problems of English language learners. Current pronunciation software tools mainly address phonemic difficulties and give little or no analytic feedback. Achievement of near native fluency however involves the ability to reproduce English prosody in terms of pitch, intensity, and duration in addition to basic phonemic competence. This paper reports on the preliminary results a project which includes phonemic diagnosis but also takes the learner forward by analysing and giving feedback on prosody. The project is being carried out in an iterative and incremental software development approach with a focus on user experience design and evaluation, market research, and with a view to developing mobile apps. The first stage of evaluation initially focussed solely on prosody and investigated 10 subjects ( 5 intermediate to advanced, 5 beginners) , whose pronunciation of a set of predetermined words, phrases and sentences were bench-marked against a set of native speaker recordings in order to establish the following: • The extent to which the learner benefited from audio only or audio plus text as a prompt to pronouncing the words, in order to establish the feasibility of training ear and vocal apparatus without text-based prompts. • To what extent the learner's pronunciation was affected by their understanding of the words they were pronouncing in order to establish the most effective learning methodology when using de-contextualised speech segments. The first finding identified the use of audio as the prime focus but some learners expressed a preference for the written word. As a result, the option of accessing the text after the audio will be included in the design of the application. The second finding indicated that for all students it was beneficial for pronunciation to understand the speech segment. Additional findings (with the beginners group) indicate a requirement to devise an adaptive system for determining any individual phonemes that learners were struggling with in one or two syllable words before moving onto analysing prosody alongside phonemic diagnosis in more complex words and phrases. Other requirements identified were the need to provide contextualisation, e.g. through illustration or dictionary support, and to cater for different levels of students by creating sets of words appropriate to their lexical knowledge. The next stage of evaluation will involve a greater number of subjects – evaluating how to present the potentially complex feedback in such a way that learners both understand and engage and are motivated to improve. A set of simple visual symbols has been devised which represent pitch, intensity, duration, and give learners instant visual feedback on all three in a simple, clear manner that avoids information overload, such as can occur with existing methods of reproducing waveforms and spectrograms. Applications of this technology include mobile apps, e.g. for single words; call centre training, e.g. of customised scripts; and a children’s adventure game e.g. to engage children in mastering English pronunciation.