Group results, to improve user cognitive load

[sharatisrani]

Current proposal is that UI will [try and] provide control to the user how they want to group. The fuller discussion (2/2/23 O&O WG) was:

Mark: Not sure either is the right solution. Fundamental problem is still the granularity in hops of up and down of ontology. I don't know how we would get this. Need a measure of significance of each plateau or tier. Some matter and some don't. Each hop you go up is less meaningful. Not just that some things are overly broad. But some things are just not significant. Systematically determining the sweet spots is difficult. Traversing by # of hops is not the best idea.
Sharat: Part of a study where it was done that way. It kind of worked. But I don't know if that should be the way to go. If anyone has an idea of how to calculate the significance of each level.
Chris: Psychological significance? I can't do it.
Mark: The only semi easy path I have only addresses part of the problem. Look at the number of X.
Chris: Enrichment calculator tries to do this.
GGlusman via chat: Enrichment analysis as Chris said sounds right.
Sharat: only
Chris Bizon via chat: Cutoffs make me itchy. We apply it to node properties and graph links as well.
From Paul Clemons (MolePro) to Everyone 03:48 PM
@chris agree re: cutoffs. Sometimes there is a principled way to set one, but more often than not they are arbitrary. We should be cautious.
Sui: I would be careful with algorithmic grouping. Need to do it on a case-by-case basis.
Tentative conclusion for MVP2 stage: simplest for user to group what they want. Andy willing to attempt such a UI.

[sharatisrani]

Today's version of this solution is to Conflate drugs, and to merge results by Subject/Object at least lightly. Much more to come past MVP2, as we get into Q's that are not inherently 1-hop answers.