Which one is more toxic? Findings from Jigsaw Rate Severity of Toxic Comments

TL;DR

This paper evaluates various transformer and traditional models for toxicity severity prediction in online comments, highlighting challenges in absolute toxicity labeling and analyzing model explanations.

Contribution

It provides a comparative analysis of models on a toxicity severity dataset and discusses issues with model predictions using explainability techniques.

Findings

Transformers outperform traditional models in toxicity severity prediction

Model predictions exhibit interpretability issues

Regression approaches are more suitable than classification for toxicity severity

Abstract

The proliferation of online hate speech has necessitated the creation of algorithms which can detect toxicity. Most of the past research focuses on this detection as a classification task, but assigning an absolute toxicity label is often tricky. Hence, few of the past works transform the same task into a regression. This paper shows the comparative evaluation of different transformers and traditional machine learning models on a recently released toxicity severity measurement dataset by Jigsaw. We further demonstrate the issues with the model predictions using explainability analysis.