None of this is as feasible as people think it is, because this is a deep problem. What about new terms like kraken ? In case of online content moderation thi

None of this is as feasible as people think it is, because this is a deep problem. What about new terms like kraken ? In case of online content moderation think in terms of signal/noise and context. The same thing is benign or malicious based on the context It finds itself in. Your filter list assumes a static context - your particular imagining of context in this case. Obviously What’s noise for you is signal/hurtful to someone else. Run this over N sub groups of subscribers and you have an impossible problem.