Fighting Offensive Language on Social Media with Unsupervised Text Style Transfer
Cícero Nogueira dos Santos, Igor V. Melnyk, Inkit Padhi · 2018
We introduce a new approach to tackle the problem of offensive language in online social media.Our approach uses unsupervised text style transfer to translate offensive sentences into non-offensive ones.We propose a new method for training encoderdecoders using non-parallel data that combines a collaborative classifier, attention and the cycle consistency loss.Experimental results on data from Twitter and Reddit show that our method outperforms a state-of-the-art text style transfer system in two out of three quantitative metrics and produces reliable non-offensive transferred sentences.