Kakao Applies 'SafeBot'... 64% Reduction in Abusive Comments

[Asia Economy Reporter Seungjin Lee] Kakao disclosed the performance analysis results of ‘SafeBot,’ an AI-based comment filtering feature, on the 14th.

Since the introduction of SafeBot in December 2020, the proportion of comments containing profanity or vulgar language and the proportion of profanity comments reported by users have significantly decreased, leading to a healthier internet comment culture.

SafeBot is a function that automatically filters comments containing profanity, vulgar language, or those violating post operation policies by analyzing them with AI technology. It replaces profanity and vulgar words with musical notes, hides comments that may cause discomfort to other users, and automatically completes the reporting process.

Comparing the average monthly figures from the second half of 2020 before SafeBot was applied with those of 2021 and 2022 after its application, comments containing profanity or vulgar language decreased to about one-third. The number of comments with profanity automatically replaced by musical notes decreased by 53.7% in 2021 and 63.8% in 2022 compared to the second half of 2020.

The proportion of profanity comments reported by users among all comments also dropped from 4.2% in the second half of 2020 to 2.4% in 2022. Additionally, profanity comments deleted due to user reports decreased by 73.6% in 2021 and 91.7% in 2022, reducing to about one-twelfth.

This is analyzed as a result of the proactive measures of SafeBot, leading to the gradual maturation of comment culture. As SafeBot swiftly filters comments containing profanity or vulgar language, users voluntarily participate in fostering a healthy comment culture. Consequently, the healthiness of the comment public forum appears to have improved.

Kakao has been striving to create a healthy comment ecosystem using AI with the goal of solving social issues through technology. As part of this, it was the first in the industry to apply the ‘profanity musical note replacement function’ in July 2017. In December 2020, it piloted the more advanced SafeBot feature and officially introduced it in December 2021. It is currently applied to the news comment services on the Daum portal and KakaoTalk View tab.

Kakao plans to further enhance SafeBot’s performance by advancing AI technology. In August, it began producing a ‘Coding Book’ to respond to hate speech, building training data to identify hate speech, which will be applied to SafeBot in the first half of next year.

Jonghan Kim, Head of Kakao Daum Business Development Office, said, “Kakao’s AI technology is analyzed to have improved the healthiness of the online public forum in the news comment service. We will continue to strive to advance related technologies so that users can use the service safely and comfortably.”

This content was produced with the assistance of AI translation services.