come up with value for 'common-ness' of a node
Opened this issue · 3 comments
We have a lot of COPs that go through something like "immune system" or some other high-level, not terribly useful concept. NAGA sort of normalizes this, but only with in a single graph. We should think about a way to derive an importance metric from hitting the sources themselves and see if we can discover which nodes are overrepresented in them.
Lets use our existing large knowledge graph to compute counts of publications outgoing from any given node. I think this ought to do the trick.
This is a good step. Another thing we could try would be doing that analysis in chemotext for nodes that we could turn into mesh terms. But I am still wondering if there is a way to get this at the KS API level. It's probably not, at least for every source.