Machine Learning Suites for Online Toxicity Detection

Abstract

To identify and classify toxic online commentary, the modern tools of data science transform raw text into key features from which either thresholding or learning algorithms can make predictions for monitoring offensive conversations. We systematically evaluate 62 classifiers representing 19 major algorithmic families against features extracted from the Jigsaw dataset of Wikipedia comments. We compare the classifiers based on statistically significant differences in accuracy and relative execution time. Among these classifiers for identifying toxic comments, tree-based algorithms provide the most transparently explainable rules and rank-order the predictive contribution of each feature. Among 28 features of syntax, sentiment, emotion and outlier word dictionaries, a simple bad word list proves most predictive of offensive commentary.

References in corpus (3)

• Deceiving Google's Perspective API Built for Detecting Toxic Comments
• Convolutional Neural Networks for Toxic Comment Classification
• Is preprocessing of text really worth your time for online comment classification?

Cited by in corpus (7)

• Defining and Detecting Toxicity on Social Media: Context and Knowledge are Key
• Empirical Analysis of Multi-Task Learning for Reducing Model Bias in Toxic Comment Detection
• Countering Online Hate Speech: An NLP Perspective
• Offensive Language Identification in Low-resourced Code-mixed Dravidian languages using Pseudo-labeling
• Identification of Bias Against People with Disabilities in Sentiment Analysis and Toxicity Detection Models
• UoB at SemEval-2021 Task 5: Extending Pre-Trained Language Models to Include Task and Domain-Specific Information for Toxic Span Prediction
• BERT-Beta: A Proactive Probabilistic Approach to Text Moderation