As more people turn online for news, there are more opportunities and a wider platform to share unregulated comments, which may lead to more instances of personal attacks and verbal abuse. Our study aims to classify hate speech in news comments and construct a dictionary of Korean hate speech. We used comments from daily ranking news in four sections of the news portal. We build a Korean hate speech dictionary that uses a word embedding technique, and we implement the classification of hate speech through various machine learning algorithms comparison. Then we present the optimal model by comparing bag-of-words and dictionary based methods. The results of this study can be used as a fundamental study for identifying Korean hate speech.