A Semi-Supervised Approach to Detect Toxic Comments

This repository contains the source code of a toxic comment detector written in Portuguese.

Requirements

• Python 3 (or later)
• pip install -r requirements.txt
• sh download.sh to download pre-trained embeddings

Reproduce the results

python main.py

References

@inproceedings{ghivvago-et-al-21,
  title = "{A} {S}emi-{S}upervised {A}pproach to {D}etect {T}oxic {C}omments",
  author = "Saraiva, Ghivvago D and Anchiêta, Rafael T and Neto, Franciso A R and Moura, Raimundo S",
  booktitle = "Proceedings of the International Conference on Recents Advances in Natural Language Processing (RANLP)",
  year = "2021",
  address = "Online",
  month = sep,
  pages = "1265--1271",
  publisher = "INCOMA Ltd.",
}