One approach for querying knowledge graph is to use nearest neighbor (NN) data structures. Read the paper On Integrating Knowledge Graph Embedding into SPARQL Query Processing for more details. Roughly speaking, the authors redesign the loss function used for knowledge graph embedding methods to generate vector representations that improve the NN search performance. For example:
Nonetheless, the following fact is know:
It is known that using the same embedding space to represent both entities and relations is less competitive compared to considering two separate spaces. (source: LogicENN: A Neural Based Knowledge Graphs Embedding Model with Logical Rules).
My question is: Since there are embedding which represent entities and relations in different spaces, how can I use NN methods to ask a Knowledge Graph?
