Implement ID clique merging about kgx

[cmungall]

@kshefchek will decide on strategy, we'll decide who to assign after that

[cmungall]

not sure I follow, 3 is the only one that uses reasoning (albeit trivial transitivity/symmetry of equivalence relation)

1 and 2 would be dumb mapping approaches. No need for external machinery. Would also allow people more control. E.g. if we don't have an equivalence axiom, could choose to map to closest concept (eg ICD10 to MONDO)

[kshefchek]

+1

[andrewsu]

@cmungall I agree with your assessment of reasoning vs dumb mapping, but it's not clear to me if the decision has been made on the preferred path moving forward (and whether that differs between an optimal long-term vs short-term hackathon/Translator plan).

Also, in this context, is a "source" a heterogeneous knowledge graph (like the ones in https://docs.google.com/spreadsheets/d/1T7qXUF0LW2ubDJtR5qWei4zRg892EMRjs8lucGOv99I/edit#gid=0), or an individual knowledge source with a focused set of edge types (like most of the ones in https://docs.google.com/spreadsheets/d/160Vzcgk5eGjtqbrKZzCyyJuPRKQV_zpk4BpZRMC70PA/edit#gid=2024684564)?

[cmungall]

So this would not preclude a recommendation of standardization of IDs upstream, this is just about getting something workable by the hackathon, and so we can make some progress on things like merging monarch-lite and reasoner graphs, and/or semmeddb #14

Yes, here source is something like a source KG in neo4j; it could also include wikidata. Something that is already "shaped" like a KG, but where we may still have a certain amount of shear