[21:25:41] <isaacj>	 leila: question regarding sockpuppet models. my understanding is that the deep learning model is predicting whether a given user has a sockpuppet account (but not necessarily which account it is). to generate predictions about whether Account A and Account B are sockpuppets from the same person, do you just compare the model representations of every user to find which are similar?
[21:27:11] <leila>	 isaacj: I'd skip talking about the deep learning model (at least in details) because we are not planning to use that now.
[21:27:52] <leila>	 isaacj: the current model does the following "Given a user and their edits till time T, is the user operating another account (at time T)?" the other questions we discuss with the audience is for future iterations over the model.
[21:29:53] <isaacj>	 leila: thanks for the clarification. oo i'll focus on the feature engineered model. and at this point we haven't decided on how to detect accounts that are probably sockpuppets of each other (just whether the edit behavior is suspicious and looks sockpuppety and potentially worthy of further investigation)
[21:30:19] <leila>	 isaacj: exactly! and thanks!
[21:30:36] <isaacj>	 :thumbs up: