SVM Approach to Forum and Comment Moderation

Social networks, blogs, and forums bring users together to build a community usually based on a system of communication through comments. Based on the degree of complexity, these systems may or may not have a moderator whose primary purpose is to remove spam or abusive comments within these systems. When a moderator is used, it is most often a human, whose time and energy must be exerted to read each and every comment making it a tedious job for a large website. A support vector machine (SVM) approach is proposed for comment moderation. Using a training corpus obtained from the popular website Youtube.com, a support vector machine is used to classify comments as abusive or not. Baseline accuracy is found by performing 10-fold cross validation on unprocessed data. Different experiments are performed on the data by preprocessing it to find if certain variations provide a more accurate estimate.