Wikipedia documents represent a group-based editorial choice about relevant information for a topic or entity. For instance the introduction of https://en.wikipedia.org/wiki/Caffeine lists anatomical entities, chemicals, diseases, etc.
So generate a ranking benchmark where the query is (?)-[related_to]->(X) and the expected answer are any entities occuring in the introductory paragraphs of X's wikipedia page (or the whole page?)
Wikipedia documents represent a group-based editorial choice about relevant information for a topic or entity. For instance the introduction of https://en.wikipedia.org/wiki/Caffeine lists anatomical entities, chemicals, diseases, etc.
So generate a ranking benchmark where the query is (?)-[related_to]->(X) and the expected answer are any entities occuring in the introductory paragraphs of X's wikipedia page (or the whole page?)