Closed sheyvaert closed 7 years ago
I was able to confirm this works, but I'm concerned that making this dependent on lumi_science & Rosette will effectively make it unusable for SEs. We can rename it something else to avoid conflict, maybe se_classifier.py?
I'm not sure I understand. Do you mean keeping the original voting_classifier.py
but renamed to se_classifier.py
alongside the run_voting_classifier.py
that imports its utilities from the new voting_classifier repo?
An alternative is to remove the dependency on lumi_science
from the voting_classifier repo. It's only there for the delangtag
function, which is small enough (and independent of Rosette) to copy into voting_classifier/util.py
Removing the dependency on lumi_science
is probably the best way. SEs will still have to pull the voting_classifier
repo locally, but that shouldn't be a big deal.
voting_classifier.py
torun_voting_classifier.py
to avoid conflictThe test with the Fidelity COSMO_SEMANTIC_TAG subset reports 81.04% overall accuracy.
Running the code will of course require one to install the
https://github.com/LuminosoInsight/voting_classifier
code