Capturing L2 segmental mispronunciations with joint-sequence models in Computer-Aided Pronunciation Training (CAPT) | IEEE Conference Publication | IEEE Xplore