Twitch uses machine learning to moderate your stream chats

AutoMod can clean up conversations without many of the usual hassles.

Robert Galbraith / Reuters

Sure, you can already take steps to keep your Twitch chat friendly, but it's a lot of work if you don't have a team of moderators. Do you really want to watch conversations like a hawk in case someone gets around your meticulously crafted filters? You might not have to after today. Twitch is introducing an AutoMod feature that uses a mix of machine learning and natural language processing to keep "inappropriate content" out of your stream chats. It not only screens for offensive language, but can spot attempts to dodge your filters through clever uses of characters and emoji. You can even set a general filtering level to determine just how profane you'd like chat to be.

If a message does run afoul of your settings, AutoMod will hold it in a publishing queue for you or moderators. You can make exceptions, in other words, or allow messages that clearly should have gone through.

The AutoMod app is available right now for English broadcasters, and as a beta in 12 other languages (including European languages, Japanese, Korean, Russian and Turkish). No, it won't be as comprehensive as having a good set of human mods, but it could make the difference between focusing on your stream and distracting yourself when a viewer gets unruly.