Interpretable Graph Representation Learning: New Theories And Applications