Evaluating online sexism detection: A comparative study of machine learning models using the EDOS dataset

Abstract

Online sexism refers to gender-based discrimination and harassment that occurs in online spaces, such as social media platforms, online communities, and forums. Machine learning models can recognize and lessen online sexism by automatically detecting sexist content in social media posts. In this experimental analysis, we have evaluated the performance of five different machine learning models, including Logistic Regression, Gaussian Naive Bayes, Decision Tree Classifier, Support Vector Machine (SVM), and KNeighbors Classifier. Our objective was to detect online sexism using the Explainable Detection of Online Sexism (EDOS) data set. We preprocess the data set by cleansing the text data with a regular expression, removing null values, removing redundant columns, and vectorizing it with TfidfVectorizer. The results of our study indicate that the Logistic Regression, Gaussian Naive Bayes, Decision Tree Classifier, and Support Vector Machine (SVM) models are efficacious in detecting occurrences of online sexism. Nonetheless, the KNeighbors Classifier algorithm shows comparatively lower accuracy in this aspect. The present analysis highlights the capacity of machine learning models to identify instances of online sexism. The highest accuracy score we obtained for the Support Vector Machine (SVM) model is 94.64%.