Cross-Modal ASR Post-Processing System for Error Correction and Utterance Rejection