Solving the Hate Comments Problem through ML-based Visualization | 2019

Visualizing negativeness of all the comments on one news article

Interactive feedback on the negativeness of comment: negativeness > 50% (top: siren icon), negativeness > 70% (bottom: animated siren icon)

Time-delayed exposure of a possible hate comment. Users have to wait for a few seconds (3 secs) to see the negative comment (negativeness > 70%) having more dislikes than likes

With the development of online news media, a culture where many people actively share their opinions through online comments has emerged. However, the severity of hate comments has also risen, increasing the demand for improvements to online commenting culture. This paper proposes a novel news comments visualization system to improve the commenting culture. The system aims to reduce the writing and exposure of hate comments through visualization and interaction design with the help of machine learning. It helps users alert themselves to the severity of hate comments, encouraging them to avoid making such comments, and also features crowdsourced hate comment filtering where the exposure of hate comments are reduced by the crowd participation. In the experiment conducted on 100 users in their 10s to 30s, the participants provided positive responses that have statistically significant differences in comparison to an existing popular news comment system. The system can be used as an alternative to current news comment systems for better online commenting culture.

Publication

Jihyun Park and Jusub Kim, Solving the Hate Comments Problem through ML-based Visualization, Journal of Digital Contents Society, Vol. 21, No. 4, pp. 771-779, Apr. 2020