Thursday, April 11, 2019

Blizzard Using Machine Learning To Combat Abusive Chat

I've been playing online games for a long, long, looong time and I still don't know how people let other people antagonize them while playing a game. Having some random 12 year old scream insults / threats into his mic is annoying (mute) but are we really at a point where we need to use machine learning to deal with it?   

Blizzard says it utilizes a deep learning system that trains itself to understand sentence structure and context. Working in tandem with another model, the AI collects a large amount of text that has been labeled good or bad and is then fed into a machine learning algorithm so it can learn all the relationships within those texts. If this model is successful, then the AI should be able to properly categorize words and sentences it has never seen before—which is one step closer to autonomous abusive chat reporting.