A method for measuring the intelligibility and nonnativeness of phone quality in foreign language pronunciation training

The problem addressed is automatically detecting, measuring and correcting nonnative pronunciation characteristics (so-called "for-eign accents") in foreign language speech. Systemic, structural and realizational differences between L1 (native language) and L2 (target language) appear as phone insertions, deletions and substitutions. A bilingual phone recognizer using native-trained acoustic models of the learner's L1 and L2 was developed to identify insertions, deletions and substitutions of L2 phones. Recognition results are combined with knowledge of phonetics, phonology and pedagogy to show learners which phones were mispronounced and to instruct how to modify their articulatory gestures for more native-sounding speech. The degree of the learner's foreign accent is measured based on the number of alternate pronunciations the learner uses; the number decreases as learning progresses. Evaluation experiments using Japanese and Ameri-can English indicate that the system is an effective component technology for computer-aided pronunciation learning.