English

Fighting Offensive Language on Social Media with Unsupervised Text Style Transfer

Computation and Language 2018-05-22 v1 Machine Learning

Abstract

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 encoder-decoders 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.

Keywords

Cite

@article{arxiv.1805.07685,
  title  = {Fighting Offensive Language on Social Media with Unsupervised Text Style Transfer},
  author = {Cicero Nogueira dos Santos and Igor Melnyk and Inkit Padhi},
  journal= {arXiv preprint arXiv:1805.07685},
  year   = {2018}
}

Comments

ACL 2018