War Against Malicious Comments... "Swearing Has Disappeared from Comments"

End of Entertainment News Comment Service with Addition of Malicious Comment Analysis AI

[Asia Economy Reporter Jin-gyu Lee] As major portals like Naver and Kakao actively engage in the war against malicious comments, the use of abusive and hateful expressions in comments has noticeably decreased. This is thanks to measures such as ending the comment service for entertainment news despite the loss of traffic and deploying artificial intelligence (AI) technology to filter out malicious comments. The series of deaths of celebrities who suffered from malicious comments, including singer and actress Sulli (real name Choi Jin-ri) last year, has raised calls in our society to improve comment services.

According to the information technology (IT) industry on the 20th, Naver updated its 'AI Cleanbot' to analyze the context of sentences in comments using AI technology to filter out malicious comments. AI Cleanbot automatically detects comments containing abusive language and profanity and blinds them. It has been successively applied to webtoons, Juniver, sports, and news since last year.

AI Cleanbot expanded the criteria for judging malicious comments from just abusive words to the context of sentences. Even if there is no profanity, if the context of the sentence is considered insulting, it is blinded. Habitual writers of malicious comments are also automatically filtered out. Habitual writers of malicious comments recognized by AI Cleanbot are restricted from using the comment service for a certain period.

Since March, Naver has stopped the comment service for entertainment news and disclosed the history of comments users have written on news articles so far. Previously, users could freely decide whether to disclose the news comments they wrote. According to Naver Data Lab, the total number of comments on Naver decreased from 867,011 as of February 28 to 334,925 as of the previous day, and the number of comments violating regulations sharply dropped from 4,709 to 415 during the same period.

Kakao also eliminated the comment service for entertainment news in October last year and added a 'discrimination/hate' category to the comment reporting criteria in the news comment services of the portal Daum and KakaoTalk #Tab in February. Additionally, functions to manage the exposure of comment areas, such as 'covering' and 'folding,' were introduced. Following the strengthening of comment sanctions, reports of malicious comments increased, while abusive and hateful expressions decreased. In the month of March after the reorganization, the number of comment reports doubled compared to before the reorganization, and even in May, after the 21st National Assembly elections, it increased by 14% compared to before the reorganization.

As the number of reports increased, the number of deleted malicious comments also rose by 65% in March compared to before the reorganization and increased by 7% in May compared to before. Comments containing abusive language and profanity decreased by more than 20%. Kakao operates a 'profanity note substitution function' that uses AI to filter abusive language and profanity in comments and replace them with musical note symbols. After the comment reorganization, comments substituted with musical notes decreased by more than 20%.