Downweight semmed edges, handle multiple edges

[cbizon]

https://arax.ncats.io/?source=ARS&id=dd3b732f-87b5-4ace-b1f8-37ee691ad065

This is:
tacrine-process-Alzheimer

The expected answer is about #7, but #1-6 is kind of ugly. Many of the edges are coming from semmed, and if there are common terms then it will make lots of different edges. I think we should include downweighting semmed in our weights, and I also am concerned that we are considering the score to be higher just because there are lots of edges joining A and B.

Anyway, I don't like the scoring for this one.