기계학습 기반 유튜브 악플 분석: “사이버렉카”에 달린 댓글의 어휘적 특성

초록

Considering the so-called “cyber wrecker,” which spreads hatred with sensational YouTube content about celebrities, this study collected comments posted on its channels, classified malicious comments with a machine learning algorithm, and analyzed their lexical characteristics. To this end, a logistic regression model was used as the algorithm and a regularization process was applied to improve prediction performance by preventing overfitting. As a result, we found that “cyber wrecker” content produced malicious comments using proper nouns, which connoted a derogatory or insulting meaning for mocking purposes, rather than swear words or slang. Also, various linguistic variations were found in the posting of malicious comments. Based on these results, we discussed the machine learning method for detecting malicious comments and ways to overcome its limitations.

키워드