A malicious-comment detection technique on the internet using the support vector machine

Abstract

The comment system on the Internet was started to facilitate the exchange of information from one person to another online. Vicious users, however, have been taking advantage of the anonymity on the Internet, aggressively issuing comments for defamation, privacy violation, and other negative purposes. In this paper, to solve this problem, a technique of detecting malicious comments that can contribute to a sound cyber ecosystem is proposed. The proposed technique has a five-step process: data collection, Korean normalization, sentiment dictionary construction, sentiment analysis, and deduction of an equation for malicious-comment detection using SVM (support vector machine) and a training dataset. To verify the effectiveness of the proposed method, its performance was assessed using a valuation data set, and the experiment results showed 87.8% accuracy, higher than that shown in the previous relevant studies.