Not sure, but would it be useful to add the ability to reinject additional texts after initial test dataset creation to increase the size of the training set ? or to generate a completely new training set?
No, this is not ok for the general pipeline (the test set is randomly built for this reason at the begining). Maybe drop the current testset & upload a new one ? To discuss with @eollion and @jboelaert
Not sure, but would it be useful to add the ability to reinject additional texts after initial test dataset creation to increase the size of the training set ? or to generate a completely new training set?