Detecting Insults in Social Commentary
Creative Commons CC BY 4.0
This report gives an overview of the various machine learning algorithms implemented to detect certain comments that may appear insulting to another participant on a social networking platform. Feature selection was performed using n-grams, and the WEKA machine learning toolkit was used to build supervised learning clasifiers, that provided an accuracy of 82% on the test dataset. The dataset was obtained from the popular data science competition portal, Kaggle.