This paper discusses a method to improve the pronunciation quality of English speech utterances made by Japanese Native Speakers solely through the use of Artificial Neural Networks（ANN）, independent of Hidden Markov Models which have commonly been used in Speech Recognition based CALL systems. This system carries out this evaluation based on detection of phonetic errors using bilingual models （Native English and Native Japanese English）, and also utilizes phone alignment via a modified CTC loss function to improve the phonetic detection rate. This research will evaluate the extent to which Neural Networks （BiLSTMs） can be used to improved English pronunciation quality without the use of an HMM finite state machine automaton.