Automated Content moderation of URLs and Uploaded Images

[iMerica]

PySlackers is virtually public and open, which makes it vulnerable to trolls. Let's write some basic content moderation automation using Tensorflow to tackle the low hanging fruit: Messages on public channels that contain URLs or uploaded images.

Basically we need to:

• Make Slackbot listen to all messages across public channels.
• If a message includes a URL that resolves to mime type image/*, run that URL though a pre-trained NSFW classifier (like this one)
• Version 1: If content is NSFW, send a message to the admins and maybe a warning to the channel.

[mattrasband]

Some version of this seems really useful, perhaps a simple pre-filter is length of membership. While it's easy to game, a 90-day cutoff is likely enough to dissuade troll-types since the instant gratification is gone.