2020 Volume 8 Issue 3 Pages 165-175
We focused on Twitter, which is expected to become an increasingly influential tool of political communication, and sought to validate public opinion formation process theories such as the concentration of opinions and announcement effects. We analyzed tweets about the Osaka Metropolis Plan, which was the subject of a referendum in May 2015. We classified tweets and accounts based on support for or opposition to the plan, using a natural language process and latent Dirichlet allocation, and by comparison with newspaper articles. We analyzed changes in the number of posts and active accounts based on the results. We could not find that majority recognition had an impact on Twitter posts. However we investigated retweet networks using social network analysis, and found that support and opposition were clearly divided in a network structure. Moreover, we found several accounts with extremely high degree centrality and betweenness centrality, thereby indicating possible validation of an opinion leader effect and the announcement effect.