The pedagogic evaluation is intended to provide clear guidelines for use, primarily by evaluating the program's suitability in relation to the projected uses. The principle of the training paradigm CAPL was evaluated to be useful, but not enough to contain most of the elements required for pronunciation training, especially in suprasegmentals. The coursewares on intonation and accent should be also integrated into the CAPL system.
As future work, the automatic pronunciation assessment method need to be devised from phoneme to word level. For other languages such as English and Korean, its effectiveness need to be investigated. In these experiments, unfortunately, we could not collect enough non-native learners's speech so that we need to run an experiment for many non-native learners of Japanese who are from a wide range of L1 backgrounds. To evaluate the effectiveness of our CAPL system exactly, we therefore need to examine the system over a long-term period using wider databases of cross-language. Our conclusions drawn from preliminary evaluation and the experimental results say that we still have some problems to solve in order to devise and implement CAPL system that is practical as an autonomous training system, but our system suggests a new type of L2 pronunciation learning method for the next generation.
