"Tone Sounds Unusual": An Analysis of 100 Million Naver Comments

[Anchor]

A comprehensive analysis of some 100 million news comments on Naver, conducted by a joint research team from KAIST and Germany, has revealed tens of thousands of accounts suspected of foreign-linked public opinion manipulation. The findings also captured signs of these accounts fueling social conflicts regardless of political orientation, whether progressive or conservative.

TJB reporter Cho Hyeongjun has the details.

[Reporter]

Naver, South Korea's largest news portal, is estimated to receive millions of comments every day.

A joint international research team from KAIST and the Max Planck Institute in Germany conducted a comprehensive analysis of approximately 110 million comments written over a 20-year period from 2006 to last year.

The goal was to detect activities where specific countries or groups attempt to influence public opinion through online news comments.

The research team used a self-developed artificial intelligence model, and the analysis method proceeded as follows.

First, 70 foreign-linked accounts previously identified by the Institute for National Security Strategy served as the starting point.

Using AI, the research team tracked accounts directly connected to these initial accounts or those that repeatedly wrote identical or similar comments on the same articles.

Primary targets included comments that used awkward Korean phrasing or could amplify social discord through blind criticism or praise of specific countries or figures.

As a result, among 4.07 million Naver news users, approximately 24,000 accounts were identified as suspected of foreign-linked influence activities.

[Lee Won-jae / Professor, KAIST Graduate School of Culture Technology: We applied 24 very conservative criteria, such as whether they posted the same number of comments during the same period and whether they copied and pasted content.]

Characteristics of the comments written by these accounts were also uncovered. First, content containing "criticism" significantly outnumbered "praise."

Furthermore, it was confirmed that these "critical" comments targeted a wide range of political factions broadly, sparing neither progressives nor conservatives.

In addition, during election periods with sharp political confrontation such as presidential elections, the number of newly active, suspected malicious accounts increased by about 51% compared to usual times.

[Lee Won-jae / Professor, KAIST Graduate School of Culture Technology: We were able to observe a phenomenon where peaks appeared very consistently during periods such as national elections or local elections.]

The research team expects that these data analysis results will help detect organized and sophisticated online activities during periods of heightened social conflict.

(Camera Reporter: Choi Un-gi TJB, CG: Park Bo-young TJB, Footage Courtesy of: KAIST)

Cho Hyeongjun, TJB
※ Please note: This article was translated by AI and may contain errors.