Towards Interpretable Explanations for Transfer Learning in Sequential Tasks