Open softvision-sergiulogigan opened 4 years ago
We ran into a contradictory labeling situation on a valid "login" related bug where the issue was filed by a user (https://github.com/webcompat/web-bugs/issues/46318) which was not labeled with ml-probability-high and another issue filed anonymously (https://github.com/webcompat/web-bugs/issues/45304) which was labeled (ml-needsdiagnosis-false and ml-probability-high)
ml-probability-high
ml-needsdiagnosis-false
I think this is part of the feedback threshold that we put. I don't think we can ensure that all our predictions are 100% correct, but we can improve things by retraining the model periodically.
We ran into a contradictory labeling situation on a valid "login" related bug where the issue was filed by a user (https://github.com/webcompat/web-bugs/issues/46318) which was not labeled with
ml-probability-high
and another issue filed anonymously (https://github.com/webcompat/web-bugs/issues/45304) which was labeled (ml-needsdiagnosis-false
andml-probability-high
)