Kakao's AI filter reduces profanity in online news comments

South Korean tech giant Kakao Corp. has reduced swearing and negative comments about news on its online portals such as Daum and Kakao View by using Safe Bot, an artificial intelligence-powered profanity filter, the company said on Wednesday.

The AI-powered filter transfers swearing and obscene words to music notes, hides the comments that violate the portals’ regulations and reports them to its internet crime complaint department.

The Korean tech giant started applying the profanity-to-music note AI system in July 2017, conducted a pilot program of Safe Bot in December 2020 and started using it in December 2021 in earnest.

Profane comments on the news services have declined since introducing the AI filter. The average number of malicious comments transferred to music notes per month has reduced by 37.2% so far in 2022, compared with that during the second half of 2020.

The swearing that portal users reported accounted for 4.2% of all news comments in the second half of 2020. It has declined to 2.4% so far in 2022, Kakao said.

Kakao is intensifying the performance of Safe Bot by advancing AI technology. It began production of a coding book to deal with online hate speech in August and will apply the relevant data to the internet bot as early as the first half of 2023.

Write to Seung-Woo Lee at leeswoo@hankyung.com
Jihyun Kim edited this article.