We present research towards developing a pronunciation dictionary that features sensitivity to learners' native phonology, specifically designed for Koreans learning English. We envision a system that can process learners' imitation of the dictionary pronunciation and instantly provide segmental and prosodic feedback. Towards this goal, we have collected a speech corpus to address the phonological and prosodic issues of Korean learners. We utilize the SUMMIT speech recognizer to model phonological rules, automatically identifying non-native phonological phenomena. These phonological rules account for the influence of learners' native language on the target language. Provided feedback points out the non-native phonological variations detected by the recognizer to improve pronunciation. Instructions are also given on the prosodic aspects of pronunciation, based on duration and F0 cues. We evaluated the feedback mechanism by rating 222 English utterances from six Korean subjects, before and after receiving feedback. 61% of the utterances were judged as improved after feedback.