Modeling pronunciation, rhythm, and intonation for automatic assessment of speech quality in aphasia rehabilitation.