Adaptation and Augmentation: Towards Better Rescoring Strategies for Automatic Speech Recognition and Spoken Term Detection