Right now we have exactly 1 feature so it's not very urgent, but we can include the vague machinery. I propose following the same outline as in madbot.classify_insulting, so the actual model lives in the ML folder (along with its training data etc.) and we just import npsklearn objects.
Right now we have exactly 1 feature so it's not very urgent, but we can include the vague machinery. I propose following the same outline as in
madbot.classify_insulting
, so the actual model lives in theML
folder (along with its training data etc.) and we just importnp
sklearn
objects.Training data relies on having a set of trolls and non-trolls, see issue https://github.com/corcra/troll-interconnector/issues/6.