search

ranking-agent / aragorn-ranker

Exposes TRAPI functions to add literature co-occurrence edges, convert publications to edge weights, and provide scores for answers.
MIT License
0 stars 1 forks source link

Giant scores #55

Closed cbizon closed 2 years ago

cbizon commented 2 years ago

https://arax.ncats.io/index.html?r=6ae8518e-7ce6-4732-9ac1-979186a29708

This query is returning some answers that have good scores, and some that are scored 10000000000000 (or so).

kennethmorton commented 2 years ago

The answer has been download and then reuploaded to QGraph for further inspection.

https://robokop.renci.org:7080/answer/zGdj64dY

QGraph seems to handle larger messages like this much better when you need to see the full JSON details.

kennethmorton commented 2 years ago

I am still investigating but I believe we have a poorly formed matrix somewhere along the way such that numpy.trace returns something close to machine precision. We then invert this trace and get these large values.

It will take more detective work to figure out how this matrix is illformed for this particular question.

cbizon commented 2 years ago

I think that this issue is fixed, is that correct @raynCovar ?

raynCovar commented 2 years ago

Correct, should have closed it. Sorry about that.