A Method for Calculating Degree of Deviation of Posts on Social Media for Public Opinion in Flaming Prediction

This study presents a method to calculate the degree of deviation of posts on social media from a set of opinions from diverse viewpoints. It's important to prevent the post on social media from getting flamed; however, even if the post does not contain malice intent, it is very difficult to predict whether the post blows up, since it may occur due to the context-dependent reasons such that the post is not accepted by public opinion. The feature of proposed method is to calculate the degree of deviation based on the magnitude of conflict between the opinions from various viewpoints, which is constructed as external resources. This allows our method to judge how moderate the opinion will be accepted by public opinion before posting it. In this study, by the experiments using a pseudo dataset of public opinion created using a Large Language Model, we evaluate the feasibility of the proposed method.