This repo contains implementation for the paper "Bitions@DravidianLangTech-EACL2021: Ensemble of Multilingual Language Models with Pseudo Labeling for Offence Detection in Dravidian Languages" to be presented at EACL-2021
In this paper, we propose use of ensembled multilingual models (transformers) with focal loss and inverse class weights. Further, we make use of psuedo labelling to improve our results.
Our key takeaways are - * * *