Improving code-switching speech recognition with data augmentation and system combination | IEEE Conference Publication | IEEE Xplore