logo

“This comment has been hidden by Cleanbot for user protection.” Our users would often find this message in the comment sections of NAVER News. This is NAVER’s Cleanbot in action, automatically detecting malicious comments. Let’s dive into NAVER’s technology and endeavor that enables users to have meaningful interactions through news comments in a safe and healthy environment.

Industry’s first AI-driven malicious comments detection for safe user environment

In April 2019, NAVER presented Cleanbot—the industry’s first AI-powered system that automatically detects and hides malicious comments on NAVER News—to help users exchange comments in a safe, protected environment. Cleanbot has since been expanded into various categories including Sports, WEBTOON, Cafe, Open Talk, CHZZK, and Clip to detect vulgar, offensive, or derogatory language considered insulting or hateful.

​

Cleanbot to filter out ill-intended comments without explicit profanity

Cleanbot initially focused on detecting explicitly offensive or vulgar language. However, NAVER has continued to enhance its performance in detailed analysis of slang and colloquial expressions through extensive datasets of malicious comments and advance its context-based trolling prediction and detection by incorporating the “Persona Embedding” technology that learns users’ writing styles.

As a result, Cleanbot has been upgraded to factor in the overall context of a sentence to detect and filter out not only provocative, discriminatory, violent, and derogatory content but also hurtful, insulting comments that may not include explicitly offensive or vulgar languages.

​

"Cleanbot’s continuous advancement does more than blocking malicious comments; it represents our key technology to create a healthy environment for digital interactions. We aim to lay a solid foundation to protect user experience and build a safe, vibrant online community through AI-powered precision analysis."

- Lee Kyu-ho, Leader of Common comments

​

Cleanbot Observer to detect malicious comments in advance through data

​

In July 2023, NAVER launched a beta version of Cleanbot Observer that analyzes the rate of attempted malicious comments based on data from Cleanbot to automatically flag high-risk articles and notify news outlets. This allowed the respective organizations to identify the articles in question and shut down the comment sections before any malicious comments were posted.

This February, NAVER upgraded Cleanbot Observer to better detect articles that cover casualties and large-scale disasters. Efforts are currently underway to automatically disable comments when such articles are labelled with disaster-related-notices such as Reporting Guidelines for Suicide Prevention. This approach is similar to the existing system that automatically turns off comments and recommendation sticker features on articles related to suicide. Through these endeavors, NAVER aims to eliminate malicious comments and promote a healthy, respectful commenting environment.

"Cleanbot Observer is designed to prevent secondary victimization caused by malicious comments and build a safer online environment. We plan to expand the scope of targeted articles and further sophisticate the detection technology to prevent any additional harm resulting from such comments."

- Im Sang-hee, Leader of News Tool Planning

With Cleanbot and Cleanbot Observer in place, malicious comments generation and exposure rates on NAVER’s services have steadily declined. NAVER will stay committed to our responsibility to create a safe, healthy comment ecosystem by consistently enhancing our AI technologies and services that support user interactions ●

​

​

► Check out Green Internet to learn more about NAVER’s efforts to promote a healthy online culture.