Finding Interpretable Concept Spaces in Node Embeddings Using Knowledge Bases