Second Language Intelligibility Trainer

A large and growing number of the residents of the United States speak English as a secondlanguage. Opportunities for employment, promotion, and education for this population can be negativelyaffected by their inability to communicate effectively in English. Little is known, however, about thespecific phonetic and phonological factors which affect the intelligibility of non-native speech. Thisproject proposes to investigate the effects on intelligibility of the phonetic and phonologicalcharacteristics of two types of foreign-accented English (native Chinese and Spanish speakers). Thisresearch will provide the empirical basis for the development of a computer-based pronunciation trainingsystem which will utilize both speaker-dependent and speaker-independent automatic speech recognitiontechnology. Unlike other efforts to develop computer-based speech trainers, this system will provide,as feedback, a measure of speech quality which has been shown to correlate well with humanjudgements. To establish the validity' of utilizing speech-recognition technology for second-languagepronunciation training the capability of the speech recognizer to perform the relevant acousticdiscriminations will be assessed. The effectiveness of the computer-based speech training aid will thenbe evaluated through pre- and post-training intelligibility testing of non-native speech.