A fundamental task in pronunciation instruction is to arrive at a better understanding of how the set of sound system required by a particular target language is produced. The linguistic system with which we are concerned here is the Japanese sound system, which involves a set of consonants and vowels. In this chapter, we explore an automatic pronunciation instruction procedure for deriving a model of vowel and consonant articulation. This approach is justified given that no consensus exists for the most appropriate feedback methods by machine.
In a second-language (L2) learning, one of the most serious problems facing all L2 teachers who deal with pronunciation is that of carry-over. Clearly, the question of carry-over is related in part to individual learner variables - degree of motivation, sensitivity to accuracy, age, education - many of which appear to be beyond an L2 teacher's control[7]. However, many studies have shown that non-standard pronunciation patterns mainly result from the differences between the sound system of L1 and that of L2. That is, a learner's native language influences pronunciation of a target language because every language has a different inventory of sounds. Fortunately, native-like pronunciation (carry-over) can be acquired by the development of self-corrective feedback techniques and critical error monitoring strategies. We believe that an important component of an effective automatic pronunciation instruction is such detection and correction of specific production problems or common mistakes that a learner would make.
Automatic procedures for pronunciation errors have focused on instructing a learner how to acquire and produce the correct production of sound in an effective way. However, relatively little research has been conducted so far for this problem. Because up to now determining learners' misarticulation by machine has been a difficult (if not impossible) task. However, recent developments in speech recognition technology has offered new challenges and opportunities on the practical (automatic) instruction with the CAPL system. The eventual goal of our research is to develop an instruction method, using state-of-the art speech recognition technologies, which is effectively designed for non-native learners in a second language. We have also conducted experiments to verify the effectiveness of our feedback methods for pronunciation errors of non-native learners.
