liwords icon indicating copy to clipboard operation
liwords copied to clipboard

Published 20 hours ago • verified publisher icon domino14

Detect profane usernames

Open ddugovic opened this issue 3 years ago • 2 comments

[REDACTED screenshot]

ddugovic avatar Feb 18 '21 12:02 ddugovic

We will deploy moderation tools that will allow us to hide profane usernames and hopefully mitigate this problem. We have decided to make this a low priority for now since any filter we can think of will become a game of Whac-A-Mole for which we will just need to use our moderation tools to solve anyway. If there is an algorithm that can effectively filter profane usernames and sufficiently solves the Scunthorpe problem, we are not aware of it. Any insights about the problem are appreciated.

jvc56 avatar Feb 23 '21 16:02 jvc56

Fair point; there are actually multiple phases of a solution:

  1. Flag potential ToS violations for manual review (as you are going to do; and yes there is some whack-a-mole)

  2. Semi-automation for example https://www.mturk.com/ - comic https://xkcd.com/1897/ image

  3. Full automation which is infeasible - comic https://xkcd.com/1319/ image

ddugovic avatar Feb 23 '21 16:02 ddugovic